1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void netdev_set_default_ethtool_ops(struct net_device *dev, 74 const struct ethtool_ops *ops); 75 76 /* Backlog congestion levels */ 77 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 78 #define NET_RX_DROP 1 /* packet dropped */ 79 80 #define MAX_NEST_DEV 8 81 82 /* 83 * Transmit return codes: transmit return codes originate from three different 84 * namespaces: 85 * 86 * - qdisc return codes 87 * - driver transmit return codes 88 * - errno values 89 * 90 * Drivers are allowed to return any one of those in their hard_start_xmit() 91 * function. Real network devices commonly used with qdiscs should only return 92 * the driver transmit return codes though - when qdiscs are used, the actual 93 * transmission happens asynchronously, so the value is not propagated to 94 * higher layers. Virtual network devices transmit synchronously; in this case 95 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 96 * others are propagated to higher layers. 97 */ 98 99 /* qdisc ->enqueue() return codes. */ 100 #define NET_XMIT_SUCCESS 0x00 101 #define NET_XMIT_DROP 0x01 /* skb dropped */ 102 #define NET_XMIT_CN 0x02 /* congestion notification */ 103 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 104 105 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 106 * indicates that the device will soon be dropping packets, or already drops 107 * some packets of the same priority; prompting us to send less aggressively. */ 108 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 109 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 110 111 /* Driver transmit return codes */ 112 #define NETDEV_TX_MASK 0xf0 113 114 enum netdev_tx { 115 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 116 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 117 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 118 }; 119 typedef enum netdev_tx netdev_tx_t; 120 121 /* 122 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 123 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 124 */ 125 static inline bool dev_xmit_complete(int rc) 126 { 127 /* 128 * Positive cases with an skb consumed by a driver: 129 * - successful transmission (rc == NETDEV_TX_OK) 130 * - error while transmitting (rc < 0) 131 * - error while queueing to a different device (rc & NET_XMIT_MASK) 132 */ 133 if (likely(rc < NET_XMIT_MASK)) 134 return true; 135 136 return false; 137 } 138 139 /* 140 * Compute the worst-case header length according to the protocols 141 * used. 142 */ 143 144 #if defined(CONFIG_HYPERV_NET) 145 # define LL_MAX_HEADER 128 146 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 147 # if defined(CONFIG_MAC80211_MESH) 148 # define LL_MAX_HEADER 128 149 # else 150 # define LL_MAX_HEADER 96 151 # endif 152 #else 153 # define LL_MAX_HEADER 32 154 #endif 155 156 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 157 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 158 #define MAX_HEADER LL_MAX_HEADER 159 #else 160 #define MAX_HEADER (LL_MAX_HEADER + 48) 161 #endif 162 163 /* 164 * Old network device statistics. Fields are native words 165 * (unsigned long) so they can be read and written atomically. 166 */ 167 168 struct net_device_stats { 169 unsigned long rx_packets; 170 unsigned long tx_packets; 171 unsigned long rx_bytes; 172 unsigned long tx_bytes; 173 unsigned long rx_errors; 174 unsigned long tx_errors; 175 unsigned long rx_dropped; 176 unsigned long tx_dropped; 177 unsigned long multicast; 178 unsigned long collisions; 179 unsigned long rx_length_errors; 180 unsigned long rx_over_errors; 181 unsigned long rx_crc_errors; 182 unsigned long rx_frame_errors; 183 unsigned long rx_fifo_errors; 184 unsigned long rx_missed_errors; 185 unsigned long tx_aborted_errors; 186 unsigned long tx_carrier_errors; 187 unsigned long tx_fifo_errors; 188 unsigned long tx_heartbeat_errors; 189 unsigned long tx_window_errors; 190 unsigned long rx_compressed; 191 unsigned long tx_compressed; 192 }; 193 194 195 #include <linux/cache.h> 196 #include <linux/skbuff.h> 197 198 #ifdef CONFIG_RPS 199 #include <linux/static_key.h> 200 extern struct static_key_false rps_needed; 201 extern struct static_key_false rfs_needed; 202 #endif 203 204 struct neighbour; 205 struct neigh_parms; 206 struct sk_buff; 207 208 struct netdev_hw_addr { 209 struct list_head list; 210 unsigned char addr[MAX_ADDR_LEN]; 211 unsigned char type; 212 #define NETDEV_HW_ADDR_T_LAN 1 213 #define NETDEV_HW_ADDR_T_SAN 2 214 #define NETDEV_HW_ADDR_T_SLAVE 3 215 #define NETDEV_HW_ADDR_T_UNICAST 4 216 #define NETDEV_HW_ADDR_T_MULTICAST 5 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 }; 351 352 enum { 353 NAPI_STATE_SCHED, /* Poll is scheduled */ 354 NAPI_STATE_MISSED, /* reschedule a napi */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 358 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 359 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 360 }; 361 362 enum { 363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 367 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_DROP, 378 GRO_CONSUMED, 379 }; 380 typedef enum gro_result gro_result_t; 381 382 /* 383 * enum rx_handler_result - Possible return values for rx_handlers. 384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 385 * further. 386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 387 * case skb->dev was changed by rx_handler. 388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 390 * 391 * rx_handlers are functions called from inside __netif_receive_skb(), to do 392 * special processing of the skb, prior to delivery to protocol handlers. 393 * 394 * Currently, a net_device can only have a single rx_handler registered. Trying 395 * to register a second rx_handler will return -EBUSY. 396 * 397 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 398 * To unregister a rx_handler on a net_device, use 399 * netdev_rx_handler_unregister(). 400 * 401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 402 * do with the skb. 403 * 404 * If the rx_handler consumed the skb in some way, it should return 405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 406 * the skb to be delivered in some other way. 407 * 408 * If the rx_handler changed skb->dev, to divert the skb to another 409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 410 * new device will be called if it exists. 411 * 412 * If the rx_handler decides the skb should be ignored, it should return 413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 414 * are registered on exact device (ptype->dev == skb->dev). 415 * 416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 417 * delivered, it should return RX_HANDLER_PASS. 418 * 419 * A device without a registered rx_handler will behave as if rx_handler 420 * returned RX_HANDLER_PASS. 421 */ 422 423 enum rx_handler_result { 424 RX_HANDLER_CONSUMED, 425 RX_HANDLER_ANOTHER, 426 RX_HANDLER_EXACT, 427 RX_HANDLER_PASS, 428 }; 429 typedef enum rx_handler_result rx_handler_result_t; 430 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 431 432 void __napi_schedule(struct napi_struct *n); 433 void __napi_schedule_irqoff(struct napi_struct *n); 434 435 static inline bool napi_disable_pending(struct napi_struct *n) 436 { 437 return test_bit(NAPI_STATE_DISABLE, &n->state); 438 } 439 440 bool napi_schedule_prep(struct napi_struct *n); 441 442 /** 443 * napi_schedule - schedule NAPI poll 444 * @n: NAPI context 445 * 446 * Schedule NAPI poll routine to be called if it is not already 447 * running. 448 */ 449 static inline void napi_schedule(struct napi_struct *n) 450 { 451 if (napi_schedule_prep(n)) 452 __napi_schedule(n); 453 } 454 455 /** 456 * napi_schedule_irqoff - schedule NAPI poll 457 * @n: NAPI context 458 * 459 * Variant of napi_schedule(), assuming hard irqs are masked. 460 */ 461 static inline void napi_schedule_irqoff(struct napi_struct *n) 462 { 463 if (napi_schedule_prep(n)) 464 __napi_schedule_irqoff(n); 465 } 466 467 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 468 static inline bool napi_reschedule(struct napi_struct *napi) 469 { 470 if (napi_schedule_prep(napi)) { 471 __napi_schedule(napi); 472 return true; 473 } 474 return false; 475 } 476 477 bool napi_complete_done(struct napi_struct *n, int work_done); 478 /** 479 * napi_complete - NAPI processing complete 480 * @n: NAPI context 481 * 482 * Mark NAPI processing as complete. 483 * Consider using napi_complete_done() instead. 484 * Return false if device should avoid rearming interrupts. 485 */ 486 static inline bool napi_complete(struct napi_struct *n) 487 { 488 return napi_complete_done(n, 0); 489 } 490 491 /** 492 * napi_hash_del - remove a NAPI from global table 493 * @napi: NAPI context 494 * 495 * Warning: caller must observe RCU grace period 496 * before freeing memory containing @napi, if 497 * this function returns true. 498 * Note: core networking stack automatically calls it 499 * from netif_napi_del(). 500 * Drivers might want to call this helper to combine all 501 * the needed RCU grace periods into a single one. 502 */ 503 bool napi_hash_del(struct napi_struct *napi); 504 505 /** 506 * napi_disable - prevent NAPI from scheduling 507 * @n: NAPI context 508 * 509 * Stop NAPI from being scheduled on this context. 510 * Waits till any outstanding processing completes. 511 */ 512 void napi_disable(struct napi_struct *n); 513 514 /** 515 * napi_enable - enable NAPI scheduling 516 * @n: NAPI context 517 * 518 * Resume NAPI from being scheduled on this context. 519 * Must be paired with napi_disable. 520 */ 521 static inline void napi_enable(struct napi_struct *n) 522 { 523 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 524 smp_mb__before_atomic(); 525 clear_bit(NAPI_STATE_SCHED, &n->state); 526 clear_bit(NAPI_STATE_NPSVC, &n->state); 527 } 528 529 /** 530 * napi_synchronize - wait until NAPI is not running 531 * @n: NAPI context 532 * 533 * Wait until NAPI is done being scheduled on this context. 534 * Waits till any outstanding processing completes but 535 * does not disable future activations. 536 */ 537 static inline void napi_synchronize(const struct napi_struct *n) 538 { 539 if (IS_ENABLED(CONFIG_SMP)) 540 while (test_bit(NAPI_STATE_SCHED, &n->state)) 541 msleep(1); 542 else 543 barrier(); 544 } 545 546 /** 547 * napi_if_scheduled_mark_missed - if napi is running, set the 548 * NAPIF_STATE_MISSED 549 * @n: NAPI context 550 * 551 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 552 * NAPI is scheduled. 553 **/ 554 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 555 { 556 unsigned long val, new; 557 558 do { 559 val = READ_ONCE(n->state); 560 if (val & NAPIF_STATE_DISABLE) 561 return true; 562 563 if (!(val & NAPIF_STATE_SCHED)) 564 return false; 565 566 new = val | NAPIF_STATE_MISSED; 567 } while (cmpxchg(&n->state, val, new) != val); 568 569 return true; 570 } 571 572 enum netdev_queue_state_t { 573 __QUEUE_STATE_DRV_XOFF, 574 __QUEUE_STATE_STACK_XOFF, 575 __QUEUE_STATE_FROZEN, 576 }; 577 578 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 579 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 580 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 581 582 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 583 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 584 QUEUE_STATE_FROZEN) 585 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 586 QUEUE_STATE_FROZEN) 587 588 /* 589 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 590 * netif_tx_* functions below are used to manipulate this flag. The 591 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 592 * queue independently. The netif_xmit_*stopped functions below are called 593 * to check if the queue has been stopped by the driver or stack (either 594 * of the XOFF bits are set in the state). Drivers should not need to call 595 * netif_xmit*stopped functions, they should only be using netif_tx_*. 596 */ 597 598 struct netdev_queue { 599 /* 600 * read-mostly part 601 */ 602 struct net_device *dev; 603 struct Qdisc __rcu *qdisc; 604 struct Qdisc *qdisc_sleeping; 605 #ifdef CONFIG_SYSFS 606 struct kobject kobj; 607 #endif 608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 609 int numa_node; 610 #endif 611 unsigned long tx_maxrate; 612 /* 613 * Number of TX timeouts for this queue 614 * (/sys/class/net/DEV/Q/trans_timeout) 615 */ 616 unsigned long trans_timeout; 617 618 /* Subordinate device that the queue has been assigned to */ 619 struct net_device *sb_dev; 620 #ifdef CONFIG_XDP_SOCKETS 621 struct xdp_umem *umem; 622 #endif 623 /* 624 * write-mostly part 625 */ 626 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 627 int xmit_lock_owner; 628 /* 629 * Time (in jiffies) of last Tx 630 */ 631 unsigned long trans_start; 632 633 unsigned long state; 634 635 #ifdef CONFIG_BQL 636 struct dql dql; 637 #endif 638 } ____cacheline_aligned_in_smp; 639 640 extern int sysctl_fb_tunnels_only_for_init_net; 641 extern int sysctl_devconf_inherit_init_net; 642 643 static inline bool net_has_fallback_tunnels(const struct net *net) 644 { 645 return net == &init_net || 646 !IS_ENABLED(CONFIG_SYSCTL) || 647 !sysctl_fb_tunnels_only_for_init_net; 648 } 649 650 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 return q->numa_node; 654 #else 655 return NUMA_NO_NODE; 656 #endif 657 } 658 659 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 660 { 661 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 662 q->numa_node = node; 663 #endif 664 } 665 666 #ifdef CONFIG_RPS 667 /* 668 * This structure holds an RPS map which can be of variable length. The 669 * map is an array of CPUs. 670 */ 671 struct rps_map { 672 unsigned int len; 673 struct rcu_head rcu; 674 u16 cpus[]; 675 }; 676 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 677 678 /* 679 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 680 * tail pointer for that CPU's input queue at the time of last enqueue, and 681 * a hardware filter index. 682 */ 683 struct rps_dev_flow { 684 u16 cpu; 685 u16 filter; 686 unsigned int last_qtail; 687 }; 688 #define RPS_NO_FILTER 0xffff 689 690 /* 691 * The rps_dev_flow_table structure contains a table of flow mappings. 692 */ 693 struct rps_dev_flow_table { 694 unsigned int mask; 695 struct rcu_head rcu; 696 struct rps_dev_flow flows[]; 697 }; 698 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 699 ((_num) * sizeof(struct rps_dev_flow))) 700 701 /* 702 * The rps_sock_flow_table contains mappings of flows to the last CPU 703 * on which they were processed by the application (set in recvmsg). 704 * Each entry is a 32bit value. Upper part is the high-order bits 705 * of flow hash, lower part is CPU number. 706 * rps_cpu_mask is used to partition the space, depending on number of 707 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 708 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 709 * meaning we use 32-6=26 bits for the hash. 710 */ 711 struct rps_sock_flow_table { 712 u32 mask; 713 714 u32 ents[] ____cacheline_aligned_in_smp; 715 }; 716 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 717 718 #define RPS_NO_CPU 0xffff 719 720 extern u32 rps_cpu_mask; 721 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 722 723 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 724 u32 hash) 725 { 726 if (table && hash) { 727 unsigned int index = hash & table->mask; 728 u32 val = hash & ~rps_cpu_mask; 729 730 /* We only give a hint, preemption can change CPU under us */ 731 val |= raw_smp_processor_id(); 732 733 if (table->ents[index] != val) 734 table->ents[index] = val; 735 } 736 } 737 738 #ifdef CONFIG_RFS_ACCEL 739 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 740 u16 filter_id); 741 #endif 742 #endif /* CONFIG_RPS */ 743 744 /* This structure contains an instance of an RX queue. */ 745 struct netdev_rx_queue { 746 #ifdef CONFIG_RPS 747 struct rps_map __rcu *rps_map; 748 struct rps_dev_flow_table __rcu *rps_flow_table; 749 #endif 750 struct kobject kobj; 751 struct net_device *dev; 752 struct xdp_rxq_info xdp_rxq; 753 #ifdef CONFIG_XDP_SOCKETS 754 struct xdp_umem *umem; 755 #endif 756 } ____cacheline_aligned_in_smp; 757 758 /* 759 * RX queue sysfs structures and functions. 760 */ 761 struct rx_queue_attribute { 762 struct attribute attr; 763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 764 ssize_t (*store)(struct netdev_rx_queue *queue, 765 const char *buf, size_t len); 766 }; 767 768 #ifdef CONFIG_XPS 769 /* 770 * This structure holds an XPS map which can be of variable length. The 771 * map is an array of queues. 772 */ 773 struct xps_map { 774 unsigned int len; 775 unsigned int alloc_len; 776 struct rcu_head rcu; 777 u16 queues[]; 778 }; 779 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 780 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 781 - sizeof(struct xps_map)) / sizeof(u16)) 782 783 /* 784 * This structure holds all XPS maps for device. Maps are indexed by CPU. 785 */ 786 struct xps_dev_maps { 787 struct rcu_head rcu; 788 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 789 }; 790 791 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 792 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 793 794 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 795 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 796 797 #endif /* CONFIG_XPS */ 798 799 #define TC_MAX_QUEUE 16 800 #define TC_BITMASK 15 801 /* HW offloaded queuing disciplines txq count and offset maps */ 802 struct netdev_tc_txq { 803 u16 count; 804 u16 offset; 805 }; 806 807 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 808 /* 809 * This structure is to hold information about the device 810 * configured to run FCoE protocol stack. 811 */ 812 struct netdev_fcoe_hbainfo { 813 char manufacturer[64]; 814 char serial_number[64]; 815 char hardware_version[64]; 816 char driver_version[64]; 817 char optionrom_version[64]; 818 char firmware_version[64]; 819 char model[256]; 820 char model_description[256]; 821 }; 822 #endif 823 824 #define MAX_PHYS_ITEM_ID_LEN 32 825 826 /* This structure holds a unique identifier to identify some 827 * physical item (port for example) used by a netdevice. 828 */ 829 struct netdev_phys_item_id { 830 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 831 unsigned char id_len; 832 }; 833 834 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 835 struct netdev_phys_item_id *b) 836 { 837 return a->id_len == b->id_len && 838 memcmp(a->id, b->id, a->id_len) == 0; 839 } 840 841 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 842 struct sk_buff *skb, 843 struct net_device *sb_dev); 844 845 enum tc_setup_type { 846 TC_SETUP_QDISC_MQPRIO, 847 TC_SETUP_CLSU32, 848 TC_SETUP_CLSFLOWER, 849 TC_SETUP_CLSMATCHALL, 850 TC_SETUP_CLSBPF, 851 TC_SETUP_BLOCK, 852 TC_SETUP_QDISC_CBS, 853 TC_SETUP_QDISC_RED, 854 TC_SETUP_QDISC_PRIO, 855 TC_SETUP_QDISC_MQ, 856 TC_SETUP_QDISC_ETF, 857 TC_SETUP_ROOT_QDISC, 858 TC_SETUP_QDISC_GRED, 859 TC_SETUP_QDISC_TAPRIO, 860 TC_SETUP_FT, 861 TC_SETUP_QDISC_ETS, 862 TC_SETUP_QDISC_TBF, 863 TC_SETUP_QDISC_FIFO, 864 }; 865 866 /* These structures hold the attributes of bpf state that are being passed 867 * to the netdevice through the bpf op. 868 */ 869 enum bpf_netdev_command { 870 /* Set or clear a bpf program used in the earliest stages of packet 871 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 872 * is responsible for calling bpf_prog_put on any old progs that are 873 * stored. In case of error, the callee need not release the new prog 874 * reference, but on success it takes ownership and must bpf_prog_put 875 * when it is no longer used. 876 */ 877 XDP_SETUP_PROG, 878 XDP_SETUP_PROG_HW, 879 /* BPF program for offload callbacks, invoked at program load time. */ 880 BPF_OFFLOAD_MAP_ALLOC, 881 BPF_OFFLOAD_MAP_FREE, 882 XDP_SETUP_XSK_UMEM, 883 }; 884 885 struct bpf_prog_offload_ops; 886 struct netlink_ext_ack; 887 struct xdp_umem; 888 struct xdp_dev_bulk_queue; 889 struct bpf_xdp_link; 890 891 enum bpf_xdp_mode { 892 XDP_MODE_SKB = 0, 893 XDP_MODE_DRV = 1, 894 XDP_MODE_HW = 2, 895 __MAX_XDP_MODE 896 }; 897 898 struct bpf_xdp_entity { 899 struct bpf_prog *prog; 900 struct bpf_xdp_link *link; 901 }; 902 903 struct netdev_bpf { 904 enum bpf_netdev_command command; 905 union { 906 /* XDP_SETUP_PROG */ 907 struct { 908 u32 flags; 909 struct bpf_prog *prog; 910 struct netlink_ext_ack *extack; 911 }; 912 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 913 struct { 914 struct bpf_offloaded_map *offmap; 915 }; 916 /* XDP_SETUP_XSK_UMEM */ 917 struct { 918 struct xdp_umem *umem; 919 u16 queue_id; 920 } xsk; 921 }; 922 }; 923 924 /* Flags for ndo_xsk_wakeup. */ 925 #define XDP_WAKEUP_RX (1 << 0) 926 #define XDP_WAKEUP_TX (1 << 1) 927 928 #ifdef CONFIG_XFRM_OFFLOAD 929 struct xfrmdev_ops { 930 int (*xdo_dev_state_add) (struct xfrm_state *x); 931 void (*xdo_dev_state_delete) (struct xfrm_state *x); 932 void (*xdo_dev_state_free) (struct xfrm_state *x); 933 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 934 struct xfrm_state *x); 935 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 936 }; 937 #endif 938 939 struct dev_ifalias { 940 struct rcu_head rcuhead; 941 char ifalias[]; 942 }; 943 944 struct devlink; 945 struct tlsdev_ops; 946 947 struct netdev_name_node { 948 struct hlist_node hlist; 949 struct list_head list; 950 struct net_device *dev; 951 const char *name; 952 }; 953 954 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 955 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 956 957 struct netdev_net_notifier { 958 struct list_head list; 959 struct notifier_block *nb; 960 }; 961 962 /* 963 * This structure defines the management hooks for network devices. 964 * The following hooks can be defined; unless noted otherwise, they are 965 * optional and can be filled with a null pointer. 966 * 967 * int (*ndo_init)(struct net_device *dev); 968 * This function is called once when a network device is registered. 969 * The network device can use this for any late stage initialization 970 * or semantic validation. It can fail with an error code which will 971 * be propagated back to register_netdev. 972 * 973 * void (*ndo_uninit)(struct net_device *dev); 974 * This function is called when device is unregistered or when registration 975 * fails. It is not called if init fails. 976 * 977 * int (*ndo_open)(struct net_device *dev); 978 * This function is called when a network device transitions to the up 979 * state. 980 * 981 * int (*ndo_stop)(struct net_device *dev); 982 * This function is called when a network device transitions to the down 983 * state. 984 * 985 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 986 * struct net_device *dev); 987 * Called when a packet needs to be transmitted. 988 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 989 * the queue before that can happen; it's for obsolete devices and weird 990 * corner cases, but the stack really does a non-trivial amount 991 * of useless work if you return NETDEV_TX_BUSY. 992 * Required; cannot be NULL. 993 * 994 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 995 * struct net_device *dev 996 * netdev_features_t features); 997 * Called by core transmit path to determine if device is capable of 998 * performing offload operations on a given packet. This is to give 999 * the device an opportunity to implement any restrictions that cannot 1000 * be otherwise expressed by feature flags. The check is called with 1001 * the set of features that the stack has calculated and it returns 1002 * those the driver believes to be appropriate. 1003 * 1004 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1005 * struct net_device *sb_dev); 1006 * Called to decide which queue to use when device supports multiple 1007 * transmit queues. 1008 * 1009 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1010 * This function is called to allow device receiver to make 1011 * changes to configuration when multicast or promiscuous is enabled. 1012 * 1013 * void (*ndo_set_rx_mode)(struct net_device *dev); 1014 * This function is called device changes address list filtering. 1015 * If driver handles unicast address filtering, it should set 1016 * IFF_UNICAST_FLT in its priv_flags. 1017 * 1018 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1019 * This function is called when the Media Access Control address 1020 * needs to be changed. If this interface is not defined, the 1021 * MAC address can not be changed. 1022 * 1023 * int (*ndo_validate_addr)(struct net_device *dev); 1024 * Test if Media Access Control address is valid for the device. 1025 * 1026 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1027 * Called when a user requests an ioctl which can't be handled by 1028 * the generic interface code. If not defined ioctls return 1029 * not supported error code. 1030 * 1031 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1032 * Used to set network devices bus interface parameters. This interface 1033 * is retained for legacy reasons; new devices should use the bus 1034 * interface (PCI) for low level management. 1035 * 1036 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1037 * Called when a user wants to change the Maximum Transfer Unit 1038 * of a device. 1039 * 1040 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1041 * Callback used when the transmitter has not made any progress 1042 * for dev->watchdog ticks. 1043 * 1044 * void (*ndo_get_stats64)(struct net_device *dev, 1045 * struct rtnl_link_stats64 *storage); 1046 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1047 * Called when a user wants to get the network device usage 1048 * statistics. Drivers must do one of the following: 1049 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1050 * rtnl_link_stats64 structure passed by the caller. 1051 * 2. Define @ndo_get_stats to update a net_device_stats structure 1052 * (which should normally be dev->stats) and return a pointer to 1053 * it. The structure may be changed asynchronously only if each 1054 * field is written atomically. 1055 * 3. Update dev->stats asynchronously and atomically, and define 1056 * neither operation. 1057 * 1058 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1059 * Return true if this device supports offload stats of this attr_id. 1060 * 1061 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1062 * void *attr_data) 1063 * Get statistics for offload operations by attr_id. Write it into the 1064 * attr_data pointer. 1065 * 1066 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1067 * If device supports VLAN filtering this function is called when a 1068 * VLAN id is registered. 1069 * 1070 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1071 * If device supports VLAN filtering this function is called when a 1072 * VLAN id is unregistered. 1073 * 1074 * void (*ndo_poll_controller)(struct net_device *dev); 1075 * 1076 * SR-IOV management functions. 1077 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1078 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1079 * u8 qos, __be16 proto); 1080 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1081 * int max_tx_rate); 1082 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1083 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1084 * int (*ndo_get_vf_config)(struct net_device *dev, 1085 * int vf, struct ifla_vf_info *ivf); 1086 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1087 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1088 * struct nlattr *port[]); 1089 * 1090 * Enable or disable the VF ability to query its RSS Redirection Table and 1091 * Hash Key. This is needed since on some devices VF share this information 1092 * with PF and querying it may introduce a theoretical security risk. 1093 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1094 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1095 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1096 * void *type_data); 1097 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1098 * This is always called from the stack with the rtnl lock held and netif 1099 * tx queues stopped. This allows the netdevice to perform queue 1100 * management safely. 1101 * 1102 * Fiber Channel over Ethernet (FCoE) offload functions. 1103 * int (*ndo_fcoe_enable)(struct net_device *dev); 1104 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1105 * so the underlying device can perform whatever needed configuration or 1106 * initialization to support acceleration of FCoE traffic. 1107 * 1108 * int (*ndo_fcoe_disable)(struct net_device *dev); 1109 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1110 * so the underlying device can perform whatever needed clean-ups to 1111 * stop supporting acceleration of FCoE traffic. 1112 * 1113 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1114 * struct scatterlist *sgl, unsigned int sgc); 1115 * Called when the FCoE Initiator wants to initialize an I/O that 1116 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1117 * perform necessary setup and returns 1 to indicate the device is set up 1118 * successfully to perform DDP on this I/O, otherwise this returns 0. 1119 * 1120 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1121 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1122 * indicated by the FC exchange id 'xid', so the underlying device can 1123 * clean up and reuse resources for later DDP requests. 1124 * 1125 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1126 * struct scatterlist *sgl, unsigned int sgc); 1127 * Called when the FCoE Target wants to initialize an I/O that 1128 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1129 * perform necessary setup and returns 1 to indicate the device is set up 1130 * successfully to perform DDP on this I/O, otherwise this returns 0. 1131 * 1132 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1133 * struct netdev_fcoe_hbainfo *hbainfo); 1134 * Called when the FCoE Protocol stack wants information on the underlying 1135 * device. This information is utilized by the FCoE protocol stack to 1136 * register attributes with Fiber Channel management service as per the 1137 * FC-GS Fabric Device Management Information(FDMI) specification. 1138 * 1139 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1140 * Called when the underlying device wants to override default World Wide 1141 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1142 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1143 * protocol stack to use. 1144 * 1145 * RFS acceleration. 1146 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1147 * u16 rxq_index, u32 flow_id); 1148 * Set hardware filter for RFS. rxq_index is the target queue index; 1149 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1150 * Return the filter ID on success, or a negative error code. 1151 * 1152 * Slave management functions (for bridge, bonding, etc). 1153 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1154 * Called to make another netdev an underling. 1155 * 1156 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1157 * Called to release previously enslaved netdev. 1158 * 1159 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1160 * struct sk_buff *skb, 1161 * bool all_slaves); 1162 * Get the xmit slave of master device. If all_slaves is true, function 1163 * assume all the slaves can transmit. 1164 * 1165 * Feature/offload setting functions. 1166 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1167 * netdev_features_t features); 1168 * Adjusts the requested feature flags according to device-specific 1169 * constraints, and returns the resulting flags. Must not modify 1170 * the device state. 1171 * 1172 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1173 * Called to update device configuration to new features. Passed 1174 * feature set might be less than what was returned by ndo_fix_features()). 1175 * Must return >0 or -errno if it changed dev->features itself. 1176 * 1177 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1178 * struct net_device *dev, 1179 * const unsigned char *addr, u16 vid, u16 flags, 1180 * struct netlink_ext_ack *extack); 1181 * Adds an FDB entry to dev for addr. 1182 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1183 * struct net_device *dev, 1184 * const unsigned char *addr, u16 vid) 1185 * Deletes the FDB entry from dev coresponding to addr. 1186 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1187 * struct net_device *dev, struct net_device *filter_dev, 1188 * int *idx) 1189 * Used to add FDB entries to dump requests. Implementers should add 1190 * entries to skb and update idx with the number of entries. 1191 * 1192 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1193 * u16 flags, struct netlink_ext_ack *extack) 1194 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1195 * struct net_device *dev, u32 filter_mask, 1196 * int nlflags) 1197 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1198 * u16 flags); 1199 * 1200 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1201 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1202 * which do not represent real hardware may define this to allow their 1203 * userspace components to manage their virtual carrier state. Devices 1204 * that determine carrier state from physical hardware properties (eg 1205 * network cables) or protocol-dependent mechanisms (eg 1206 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1207 * 1208 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1209 * struct netdev_phys_item_id *ppid); 1210 * Called to get ID of physical port of this device. If driver does 1211 * not implement this, it is assumed that the hw is not able to have 1212 * multiple net devices on single physical port. 1213 * 1214 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1215 * struct netdev_phys_item_id *ppid) 1216 * Called to get the parent ID of the physical port of this device. 1217 * 1218 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1219 * struct udp_tunnel_info *ti); 1220 * Called by UDP tunnel to notify a driver about the UDP port and socket 1221 * address family that a UDP tunnel is listnening to. It is called only 1222 * when a new port starts listening. The operation is protected by the 1223 * RTNL. 1224 * 1225 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1226 * struct udp_tunnel_info *ti); 1227 * Called by UDP tunnel to notify the driver about a UDP port and socket 1228 * address family that the UDP tunnel is not listening to anymore. The 1229 * operation is protected by the RTNL. 1230 * 1231 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1232 * struct net_device *dev) 1233 * Called by upper layer devices to accelerate switching or other 1234 * station functionality into hardware. 'pdev is the lowerdev 1235 * to use for the offload and 'dev' is the net device that will 1236 * back the offload. Returns a pointer to the private structure 1237 * the upper layer will maintain. 1238 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1239 * Called by upper layer device to delete the station created 1240 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1241 * the station and priv is the structure returned by the add 1242 * operation. 1243 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1244 * int queue_index, u32 maxrate); 1245 * Called when a user wants to set a max-rate limitation of specific 1246 * TX queue. 1247 * int (*ndo_get_iflink)(const struct net_device *dev); 1248 * Called to get the iflink value of this device. 1249 * void (*ndo_change_proto_down)(struct net_device *dev, 1250 * bool proto_down); 1251 * This function is used to pass protocol port error state information 1252 * to the switch driver. The switch driver can react to the proto_down 1253 * by doing a phys down on the associated switch port. 1254 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1255 * This function is used to get egress tunnel information for given skb. 1256 * This is useful for retrieving outer tunnel header parameters while 1257 * sampling packet. 1258 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1259 * This function is used to specify the headroom that the skb must 1260 * consider when allocation skb during packet reception. Setting 1261 * appropriate rx headroom value allows avoiding skb head copy on 1262 * forward. Setting a negative value resets the rx headroom to the 1263 * default value. 1264 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1265 * This function is used to set or query state related to XDP on the 1266 * netdevice and manage BPF offload. See definition of 1267 * enum bpf_netdev_command for details. 1268 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1269 * u32 flags); 1270 * This function is used to submit @n XDP packets for transmit on a 1271 * netdevice. Returns number of frames successfully transmitted, frames 1272 * that got dropped are freed/returned via xdp_return_frame(). 1273 * Returns negative number, means general error invoking ndo, meaning 1274 * no frames were xmit'ed and core-caller will free all frames. 1275 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1276 * This function is used to wake up the softirq, ksoftirqd or kthread 1277 * responsible for sending and/or receiving packets on a specific 1278 * queue id bound to an AF_XDP socket. The flags field specifies if 1279 * only RX, only Tx, or both should be woken up using the flags 1280 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1281 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1282 * Get devlink port instance associated with a given netdev. 1283 * Called with a reference on the netdevice and devlink locks only, 1284 * rtnl_lock is not held. 1285 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1286 * int cmd); 1287 * Add, change, delete or get information on an IPv4 tunnel. 1288 */ 1289 struct net_device_ops { 1290 int (*ndo_init)(struct net_device *dev); 1291 void (*ndo_uninit)(struct net_device *dev); 1292 int (*ndo_open)(struct net_device *dev); 1293 int (*ndo_stop)(struct net_device *dev); 1294 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1295 struct net_device *dev); 1296 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1297 struct net_device *dev, 1298 netdev_features_t features); 1299 u16 (*ndo_select_queue)(struct net_device *dev, 1300 struct sk_buff *skb, 1301 struct net_device *sb_dev); 1302 void (*ndo_change_rx_flags)(struct net_device *dev, 1303 int flags); 1304 void (*ndo_set_rx_mode)(struct net_device *dev); 1305 int (*ndo_set_mac_address)(struct net_device *dev, 1306 void *addr); 1307 int (*ndo_validate_addr)(struct net_device *dev); 1308 int (*ndo_do_ioctl)(struct net_device *dev, 1309 struct ifreq *ifr, int cmd); 1310 int (*ndo_set_config)(struct net_device *dev, 1311 struct ifmap *map); 1312 int (*ndo_change_mtu)(struct net_device *dev, 1313 int new_mtu); 1314 int (*ndo_neigh_setup)(struct net_device *dev, 1315 struct neigh_parms *); 1316 void (*ndo_tx_timeout) (struct net_device *dev, 1317 unsigned int txqueue); 1318 1319 void (*ndo_get_stats64)(struct net_device *dev, 1320 struct rtnl_link_stats64 *storage); 1321 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1322 int (*ndo_get_offload_stats)(int attr_id, 1323 const struct net_device *dev, 1324 void *attr_data); 1325 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1326 1327 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1328 __be16 proto, u16 vid); 1329 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1330 __be16 proto, u16 vid); 1331 #ifdef CONFIG_NET_POLL_CONTROLLER 1332 void (*ndo_poll_controller)(struct net_device *dev); 1333 int (*ndo_netpoll_setup)(struct net_device *dev, 1334 struct netpoll_info *info); 1335 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1336 #endif 1337 int (*ndo_set_vf_mac)(struct net_device *dev, 1338 int queue, u8 *mac); 1339 int (*ndo_set_vf_vlan)(struct net_device *dev, 1340 int queue, u16 vlan, 1341 u8 qos, __be16 proto); 1342 int (*ndo_set_vf_rate)(struct net_device *dev, 1343 int vf, int min_tx_rate, 1344 int max_tx_rate); 1345 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1346 int vf, bool setting); 1347 int (*ndo_set_vf_trust)(struct net_device *dev, 1348 int vf, bool setting); 1349 int (*ndo_get_vf_config)(struct net_device *dev, 1350 int vf, 1351 struct ifla_vf_info *ivf); 1352 int (*ndo_set_vf_link_state)(struct net_device *dev, 1353 int vf, int link_state); 1354 int (*ndo_get_vf_stats)(struct net_device *dev, 1355 int vf, 1356 struct ifla_vf_stats 1357 *vf_stats); 1358 int (*ndo_set_vf_port)(struct net_device *dev, 1359 int vf, 1360 struct nlattr *port[]); 1361 int (*ndo_get_vf_port)(struct net_device *dev, 1362 int vf, struct sk_buff *skb); 1363 int (*ndo_get_vf_guid)(struct net_device *dev, 1364 int vf, 1365 struct ifla_vf_guid *node_guid, 1366 struct ifla_vf_guid *port_guid); 1367 int (*ndo_set_vf_guid)(struct net_device *dev, 1368 int vf, u64 guid, 1369 int guid_type); 1370 int (*ndo_set_vf_rss_query_en)( 1371 struct net_device *dev, 1372 int vf, bool setting); 1373 int (*ndo_setup_tc)(struct net_device *dev, 1374 enum tc_setup_type type, 1375 void *type_data); 1376 #if IS_ENABLED(CONFIG_FCOE) 1377 int (*ndo_fcoe_enable)(struct net_device *dev); 1378 int (*ndo_fcoe_disable)(struct net_device *dev); 1379 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1380 u16 xid, 1381 struct scatterlist *sgl, 1382 unsigned int sgc); 1383 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1384 u16 xid); 1385 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1386 u16 xid, 1387 struct scatterlist *sgl, 1388 unsigned int sgc); 1389 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1390 struct netdev_fcoe_hbainfo *hbainfo); 1391 #endif 1392 1393 #if IS_ENABLED(CONFIG_LIBFCOE) 1394 #define NETDEV_FCOE_WWNN 0 1395 #define NETDEV_FCOE_WWPN 1 1396 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1397 u64 *wwn, int type); 1398 #endif 1399 1400 #ifdef CONFIG_RFS_ACCEL 1401 int (*ndo_rx_flow_steer)(struct net_device *dev, 1402 const struct sk_buff *skb, 1403 u16 rxq_index, 1404 u32 flow_id); 1405 #endif 1406 int (*ndo_add_slave)(struct net_device *dev, 1407 struct net_device *slave_dev, 1408 struct netlink_ext_ack *extack); 1409 int (*ndo_del_slave)(struct net_device *dev, 1410 struct net_device *slave_dev); 1411 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1412 struct sk_buff *skb, 1413 bool all_slaves); 1414 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1415 netdev_features_t features); 1416 int (*ndo_set_features)(struct net_device *dev, 1417 netdev_features_t features); 1418 int (*ndo_neigh_construct)(struct net_device *dev, 1419 struct neighbour *n); 1420 void (*ndo_neigh_destroy)(struct net_device *dev, 1421 struct neighbour *n); 1422 1423 int (*ndo_fdb_add)(struct ndmsg *ndm, 1424 struct nlattr *tb[], 1425 struct net_device *dev, 1426 const unsigned char *addr, 1427 u16 vid, 1428 u16 flags, 1429 struct netlink_ext_ack *extack); 1430 int (*ndo_fdb_del)(struct ndmsg *ndm, 1431 struct nlattr *tb[], 1432 struct net_device *dev, 1433 const unsigned char *addr, 1434 u16 vid); 1435 int (*ndo_fdb_dump)(struct sk_buff *skb, 1436 struct netlink_callback *cb, 1437 struct net_device *dev, 1438 struct net_device *filter_dev, 1439 int *idx); 1440 int (*ndo_fdb_get)(struct sk_buff *skb, 1441 struct nlattr *tb[], 1442 struct net_device *dev, 1443 const unsigned char *addr, 1444 u16 vid, u32 portid, u32 seq, 1445 struct netlink_ext_ack *extack); 1446 int (*ndo_bridge_setlink)(struct net_device *dev, 1447 struct nlmsghdr *nlh, 1448 u16 flags, 1449 struct netlink_ext_ack *extack); 1450 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1451 u32 pid, u32 seq, 1452 struct net_device *dev, 1453 u32 filter_mask, 1454 int nlflags); 1455 int (*ndo_bridge_dellink)(struct net_device *dev, 1456 struct nlmsghdr *nlh, 1457 u16 flags); 1458 int (*ndo_change_carrier)(struct net_device *dev, 1459 bool new_carrier); 1460 int (*ndo_get_phys_port_id)(struct net_device *dev, 1461 struct netdev_phys_item_id *ppid); 1462 int (*ndo_get_port_parent_id)(struct net_device *dev, 1463 struct netdev_phys_item_id *ppid); 1464 int (*ndo_get_phys_port_name)(struct net_device *dev, 1465 char *name, size_t len); 1466 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1467 struct udp_tunnel_info *ti); 1468 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1469 struct udp_tunnel_info *ti); 1470 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1471 struct net_device *dev); 1472 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1473 void *priv); 1474 1475 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1476 int queue_index, 1477 u32 maxrate); 1478 int (*ndo_get_iflink)(const struct net_device *dev); 1479 int (*ndo_change_proto_down)(struct net_device *dev, 1480 bool proto_down); 1481 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1482 struct sk_buff *skb); 1483 void (*ndo_set_rx_headroom)(struct net_device *dev, 1484 int needed_headroom); 1485 int (*ndo_bpf)(struct net_device *dev, 1486 struct netdev_bpf *bpf); 1487 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1488 struct xdp_frame **xdp, 1489 u32 flags); 1490 int (*ndo_xsk_wakeup)(struct net_device *dev, 1491 u32 queue_id, u32 flags); 1492 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1493 int (*ndo_tunnel_ctl)(struct net_device *dev, 1494 struct ip_tunnel_parm *p, int cmd); 1495 }; 1496 1497 /** 1498 * enum net_device_priv_flags - &struct net_device priv_flags 1499 * 1500 * These are the &struct net_device, they are only set internally 1501 * by drivers and used in the kernel. These flags are invisible to 1502 * userspace; this means that the order of these flags can change 1503 * during any kernel release. 1504 * 1505 * You should have a pretty good reason to be extending these flags. 1506 * 1507 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1508 * @IFF_EBRIDGE: Ethernet bridging device 1509 * @IFF_BONDING: bonding master or slave 1510 * @IFF_ISATAP: ISATAP interface (RFC4214) 1511 * @IFF_WAN_HDLC: WAN HDLC device 1512 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1513 * release skb->dst 1514 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1515 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1516 * @IFF_MACVLAN_PORT: device used as macvlan port 1517 * @IFF_BRIDGE_PORT: device used as bridge port 1518 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1519 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1520 * @IFF_UNICAST_FLT: Supports unicast filtering 1521 * @IFF_TEAM_PORT: device used as team port 1522 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1523 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1524 * change when it's running 1525 * @IFF_MACVLAN: Macvlan device 1526 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1527 * underlying stacked devices 1528 * @IFF_L3MDEV_MASTER: device is an L3 master device 1529 * @IFF_NO_QUEUE: device can run without qdisc attached 1530 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1531 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1532 * @IFF_TEAM: device is a team device 1533 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1534 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1535 * entity (i.e. the master device for bridged veth) 1536 * @IFF_MACSEC: device is a MACsec device 1537 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1538 * @IFF_FAILOVER: device is a failover master device 1539 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1540 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1541 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1542 */ 1543 enum netdev_priv_flags { 1544 IFF_802_1Q_VLAN = 1<<0, 1545 IFF_EBRIDGE = 1<<1, 1546 IFF_BONDING = 1<<2, 1547 IFF_ISATAP = 1<<3, 1548 IFF_WAN_HDLC = 1<<4, 1549 IFF_XMIT_DST_RELEASE = 1<<5, 1550 IFF_DONT_BRIDGE = 1<<6, 1551 IFF_DISABLE_NETPOLL = 1<<7, 1552 IFF_MACVLAN_PORT = 1<<8, 1553 IFF_BRIDGE_PORT = 1<<9, 1554 IFF_OVS_DATAPATH = 1<<10, 1555 IFF_TX_SKB_SHARING = 1<<11, 1556 IFF_UNICAST_FLT = 1<<12, 1557 IFF_TEAM_PORT = 1<<13, 1558 IFF_SUPP_NOFCS = 1<<14, 1559 IFF_LIVE_ADDR_CHANGE = 1<<15, 1560 IFF_MACVLAN = 1<<16, 1561 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1562 IFF_L3MDEV_MASTER = 1<<18, 1563 IFF_NO_QUEUE = 1<<19, 1564 IFF_OPENVSWITCH = 1<<20, 1565 IFF_L3MDEV_SLAVE = 1<<21, 1566 IFF_TEAM = 1<<22, 1567 IFF_RXFH_CONFIGURED = 1<<23, 1568 IFF_PHONY_HEADROOM = 1<<24, 1569 IFF_MACSEC = 1<<25, 1570 IFF_NO_RX_HANDLER = 1<<26, 1571 IFF_FAILOVER = 1<<27, 1572 IFF_FAILOVER_SLAVE = 1<<28, 1573 IFF_L3MDEV_RX_HANDLER = 1<<29, 1574 IFF_LIVE_RENAME_OK = 1<<30, 1575 }; 1576 1577 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1578 #define IFF_EBRIDGE IFF_EBRIDGE 1579 #define IFF_BONDING IFF_BONDING 1580 #define IFF_ISATAP IFF_ISATAP 1581 #define IFF_WAN_HDLC IFF_WAN_HDLC 1582 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1583 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1584 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1585 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1586 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1587 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1588 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1589 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1590 #define IFF_TEAM_PORT IFF_TEAM_PORT 1591 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1592 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1593 #define IFF_MACVLAN IFF_MACVLAN 1594 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1595 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1596 #define IFF_NO_QUEUE IFF_NO_QUEUE 1597 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1598 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1599 #define IFF_TEAM IFF_TEAM 1600 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1601 #define IFF_MACSEC IFF_MACSEC 1602 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1603 #define IFF_FAILOVER IFF_FAILOVER 1604 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1605 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1606 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1607 1608 /** 1609 * struct net_device - The DEVICE structure. 1610 * 1611 * Actually, this whole structure is a big mistake. It mixes I/O 1612 * data with strictly "high-level" data, and it has to know about 1613 * almost every data structure used in the INET module. 1614 * 1615 * @name: This is the first field of the "visible" part of this structure 1616 * (i.e. as seen by users in the "Space.c" file). It is the name 1617 * of the interface. 1618 * 1619 * @name_node: Name hashlist node 1620 * @ifalias: SNMP alias 1621 * @mem_end: Shared memory end 1622 * @mem_start: Shared memory start 1623 * @base_addr: Device I/O address 1624 * @irq: Device IRQ number 1625 * 1626 * @state: Generic network queuing layer state, see netdev_state_t 1627 * @dev_list: The global list of network devices 1628 * @napi_list: List entry used for polling NAPI devices 1629 * @unreg_list: List entry when we are unregistering the 1630 * device; see the function unregister_netdev 1631 * @close_list: List entry used when we are closing the device 1632 * @ptype_all: Device-specific packet handlers for all protocols 1633 * @ptype_specific: Device-specific, protocol-specific packet handlers 1634 * 1635 * @adj_list: Directly linked devices, like slaves for bonding 1636 * @features: Currently active device features 1637 * @hw_features: User-changeable features 1638 * 1639 * @wanted_features: User-requested features 1640 * @vlan_features: Mask of features inheritable by VLAN devices 1641 * 1642 * @hw_enc_features: Mask of features inherited by encapsulating devices 1643 * This field indicates what encapsulation 1644 * offloads the hardware is capable of doing, 1645 * and drivers will need to set them appropriately. 1646 * 1647 * @mpls_features: Mask of features inheritable by MPLS 1648 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1649 * 1650 * @ifindex: interface index 1651 * @group: The group the device belongs to 1652 * 1653 * @stats: Statistics struct, which was left as a legacy, use 1654 * rtnl_link_stats64 instead 1655 * 1656 * @rx_dropped: Dropped packets by core network, 1657 * do not use this in drivers 1658 * @tx_dropped: Dropped packets by core network, 1659 * do not use this in drivers 1660 * @rx_nohandler: nohandler dropped packets by core network on 1661 * inactive devices, do not use this in drivers 1662 * @carrier_up_count: Number of times the carrier has been up 1663 * @carrier_down_count: Number of times the carrier has been down 1664 * 1665 * @wireless_handlers: List of functions to handle Wireless Extensions, 1666 * instead of ioctl, 1667 * see <net/iw_handler.h> for details. 1668 * @wireless_data: Instance data managed by the core of wireless extensions 1669 * 1670 * @netdev_ops: Includes several pointers to callbacks, 1671 * if one wants to override the ndo_*() functions 1672 * @ethtool_ops: Management operations 1673 * @l3mdev_ops: Layer 3 master device operations 1674 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1675 * discovery handling. Necessary for e.g. 6LoWPAN. 1676 * @xfrmdev_ops: Transformation offload operations 1677 * @tlsdev_ops: Transport Layer Security offload operations 1678 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1679 * of Layer 2 headers. 1680 * 1681 * @flags: Interface flags (a la BSD) 1682 * @priv_flags: Like 'flags' but invisible to userspace, 1683 * see if.h for the definitions 1684 * @gflags: Global flags ( kept as legacy ) 1685 * @padded: How much padding added by alloc_netdev() 1686 * @operstate: RFC2863 operstate 1687 * @link_mode: Mapping policy to operstate 1688 * @if_port: Selectable AUI, TP, ... 1689 * @dma: DMA channel 1690 * @mtu: Interface MTU value 1691 * @min_mtu: Interface Minimum MTU value 1692 * @max_mtu: Interface Maximum MTU value 1693 * @type: Interface hardware type 1694 * @hard_header_len: Maximum hardware header length. 1695 * @min_header_len: Minimum hardware header length 1696 * 1697 * @needed_headroom: Extra headroom the hardware may need, but not in all 1698 * cases can this be guaranteed 1699 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1700 * cases can this be guaranteed. Some cases also use 1701 * LL_MAX_HEADER instead to allocate the skb 1702 * 1703 * interface address info: 1704 * 1705 * @perm_addr: Permanent hw address 1706 * @addr_assign_type: Hw address assignment type 1707 * @addr_len: Hardware address length 1708 * @upper_level: Maximum depth level of upper devices. 1709 * @lower_level: Maximum depth level of lower devices. 1710 * @neigh_priv_len: Used in neigh_alloc() 1711 * @dev_id: Used to differentiate devices that share 1712 * the same link layer address 1713 * @dev_port: Used to differentiate devices that share 1714 * the same function 1715 * @addr_list_lock: XXX: need comments on this one 1716 * @name_assign_type: network interface name assignment type 1717 * @uc_promisc: Counter that indicates promiscuous mode 1718 * has been enabled due to the need to listen to 1719 * additional unicast addresses in a device that 1720 * does not implement ndo_set_rx_mode() 1721 * @uc: unicast mac addresses 1722 * @mc: multicast mac addresses 1723 * @dev_addrs: list of device hw addresses 1724 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1725 * @promiscuity: Number of times the NIC is told to work in 1726 * promiscuous mode; if it becomes 0 the NIC will 1727 * exit promiscuous mode 1728 * @allmulti: Counter, enables or disables allmulticast mode 1729 * 1730 * @vlan_info: VLAN info 1731 * @dsa_ptr: dsa specific data 1732 * @tipc_ptr: TIPC specific data 1733 * @atalk_ptr: AppleTalk link 1734 * @ip_ptr: IPv4 specific data 1735 * @dn_ptr: DECnet specific data 1736 * @ip6_ptr: IPv6 specific data 1737 * @ax25_ptr: AX.25 specific data 1738 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1739 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1740 * device struct 1741 * @mpls_ptr: mpls_dev struct pointer 1742 * 1743 * @dev_addr: Hw address (before bcast, 1744 * because most packets are unicast) 1745 * 1746 * @_rx: Array of RX queues 1747 * @num_rx_queues: Number of RX queues 1748 * allocated at register_netdev() time 1749 * @real_num_rx_queues: Number of RX queues currently active in device 1750 * @xdp_prog: XDP sockets filter program pointer 1751 * @gro_flush_timeout: timeout for GRO layer in NAPI 1752 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1753 * allow to avoid NIC hard IRQ, on busy queues. 1754 * 1755 * @rx_handler: handler for received packets 1756 * @rx_handler_data: XXX: need comments on this one 1757 * @miniq_ingress: ingress/clsact qdisc specific data for 1758 * ingress processing 1759 * @ingress_queue: XXX: need comments on this one 1760 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1761 * @broadcast: hw bcast address 1762 * 1763 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1764 * indexed by RX queue number. Assigned by driver. 1765 * This must only be set if the ndo_rx_flow_steer 1766 * operation is defined 1767 * @index_hlist: Device index hash chain 1768 * 1769 * @_tx: Array of TX queues 1770 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1771 * @real_num_tx_queues: Number of TX queues currently active in device 1772 * @qdisc: Root qdisc from userspace point of view 1773 * @tx_queue_len: Max frames per queue allowed 1774 * @tx_global_lock: XXX: need comments on this one 1775 * @xdp_bulkq: XDP device bulk queue 1776 * @xps_cpus_map: all CPUs map for XPS device 1777 * @xps_rxqs_map: all RXQs map for XPS device 1778 * 1779 * @xps_maps: XXX: need comments on this one 1780 * @miniq_egress: clsact qdisc specific data for 1781 * egress processing 1782 * @qdisc_hash: qdisc hash table 1783 * @watchdog_timeo: Represents the timeout that is used by 1784 * the watchdog (see dev_watchdog()) 1785 * @watchdog_timer: List of timers 1786 * 1787 * @proto_down_reason: reason a netdev interface is held down 1788 * @pcpu_refcnt: Number of references to this device 1789 * @todo_list: Delayed register/unregister 1790 * @link_watch_list: XXX: need comments on this one 1791 * 1792 * @reg_state: Register/unregister state machine 1793 * @dismantle: Device is going to be freed 1794 * @rtnl_link_state: This enum represents the phases of creating 1795 * a new link 1796 * 1797 * @needs_free_netdev: Should unregister perform free_netdev? 1798 * @priv_destructor: Called from unregister 1799 * @npinfo: XXX: need comments on this one 1800 * @nd_net: Network namespace this network device is inside 1801 * 1802 * @ml_priv: Mid-layer private 1803 * @lstats: Loopback statistics 1804 * @tstats: Tunnel statistics 1805 * @dstats: Dummy statistics 1806 * @vstats: Virtual ethernet statistics 1807 * 1808 * @garp_port: GARP 1809 * @mrp_port: MRP 1810 * 1811 * @dev: Class/net/name entry 1812 * @sysfs_groups: Space for optional device, statistics and wireless 1813 * sysfs groups 1814 * 1815 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1816 * @rtnl_link_ops: Rtnl_link_ops 1817 * 1818 * @gso_max_size: Maximum size of generic segmentation offload 1819 * @gso_max_segs: Maximum number of segments that can be passed to the 1820 * NIC for GSO 1821 * 1822 * @dcbnl_ops: Data Center Bridging netlink ops 1823 * @num_tc: Number of traffic classes in the net device 1824 * @tc_to_txq: XXX: need comments on this one 1825 * @prio_tc_map: XXX: need comments on this one 1826 * 1827 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1828 * 1829 * @priomap: XXX: need comments on this one 1830 * @phydev: Physical device may attach itself 1831 * for hardware timestamping 1832 * @sfp_bus: attached &struct sfp_bus structure. 1833 * 1834 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1835 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1836 * 1837 * @proto_down: protocol port state information can be sent to the 1838 * switch driver and used to set the phys state of the 1839 * switch port. 1840 * 1841 * @wol_enabled: Wake-on-LAN is enabled 1842 * 1843 * @net_notifier_list: List of per-net netdev notifier block 1844 * that follow this device when it is moved 1845 * to another network namespace. 1846 * 1847 * @macsec_ops: MACsec offloading ops 1848 * 1849 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1850 * offload capabilities of the device 1851 * @udp_tunnel_nic: UDP tunnel offload state 1852 * @xdp_state: stores info on attached XDP BPF programs 1853 * 1854 * FIXME: cleanup struct net_device such that network protocol info 1855 * moves out. 1856 */ 1857 1858 struct net_device { 1859 char name[IFNAMSIZ]; 1860 struct netdev_name_node *name_node; 1861 struct dev_ifalias __rcu *ifalias; 1862 /* 1863 * I/O specific fields 1864 * FIXME: Merge these and struct ifmap into one 1865 */ 1866 unsigned long mem_end; 1867 unsigned long mem_start; 1868 unsigned long base_addr; 1869 int irq; 1870 1871 /* 1872 * Some hardware also needs these fields (state,dev_list, 1873 * napi_list,unreg_list,close_list) but they are not 1874 * part of the usual set specified in Space.c. 1875 */ 1876 1877 unsigned long state; 1878 1879 struct list_head dev_list; 1880 struct list_head napi_list; 1881 struct list_head unreg_list; 1882 struct list_head close_list; 1883 struct list_head ptype_all; 1884 struct list_head ptype_specific; 1885 1886 struct { 1887 struct list_head upper; 1888 struct list_head lower; 1889 } adj_list; 1890 1891 netdev_features_t features; 1892 netdev_features_t hw_features; 1893 netdev_features_t wanted_features; 1894 netdev_features_t vlan_features; 1895 netdev_features_t hw_enc_features; 1896 netdev_features_t mpls_features; 1897 netdev_features_t gso_partial_features; 1898 1899 int ifindex; 1900 int group; 1901 1902 struct net_device_stats stats; 1903 1904 atomic_long_t rx_dropped; 1905 atomic_long_t tx_dropped; 1906 atomic_long_t rx_nohandler; 1907 1908 /* Stats to monitor link on/off, flapping */ 1909 atomic_t carrier_up_count; 1910 atomic_t carrier_down_count; 1911 1912 #ifdef CONFIG_WIRELESS_EXT 1913 const struct iw_handler_def *wireless_handlers; 1914 struct iw_public_data *wireless_data; 1915 #endif 1916 const struct net_device_ops *netdev_ops; 1917 const struct ethtool_ops *ethtool_ops; 1918 #ifdef CONFIG_NET_L3_MASTER_DEV 1919 const struct l3mdev_ops *l3mdev_ops; 1920 #endif 1921 #if IS_ENABLED(CONFIG_IPV6) 1922 const struct ndisc_ops *ndisc_ops; 1923 #endif 1924 1925 #ifdef CONFIG_XFRM_OFFLOAD 1926 const struct xfrmdev_ops *xfrmdev_ops; 1927 #endif 1928 1929 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1930 const struct tlsdev_ops *tlsdev_ops; 1931 #endif 1932 1933 const struct header_ops *header_ops; 1934 1935 unsigned int flags; 1936 unsigned int priv_flags; 1937 1938 unsigned short gflags; 1939 unsigned short padded; 1940 1941 unsigned char operstate; 1942 unsigned char link_mode; 1943 1944 unsigned char if_port; 1945 unsigned char dma; 1946 1947 /* Note : dev->mtu is often read without holding a lock. 1948 * Writers usually hold RTNL. 1949 * It is recommended to use READ_ONCE() to annotate the reads, 1950 * and to use WRITE_ONCE() to annotate the writes. 1951 */ 1952 unsigned int mtu; 1953 unsigned int min_mtu; 1954 unsigned int max_mtu; 1955 unsigned short type; 1956 unsigned short hard_header_len; 1957 unsigned char min_header_len; 1958 1959 unsigned short needed_headroom; 1960 unsigned short needed_tailroom; 1961 1962 /* Interface address info. */ 1963 unsigned char perm_addr[MAX_ADDR_LEN]; 1964 unsigned char addr_assign_type; 1965 unsigned char addr_len; 1966 unsigned char upper_level; 1967 unsigned char lower_level; 1968 unsigned short neigh_priv_len; 1969 unsigned short dev_id; 1970 unsigned short dev_port; 1971 spinlock_t addr_list_lock; 1972 unsigned char name_assign_type; 1973 bool uc_promisc; 1974 struct netdev_hw_addr_list uc; 1975 struct netdev_hw_addr_list mc; 1976 struct netdev_hw_addr_list dev_addrs; 1977 1978 #ifdef CONFIG_SYSFS 1979 struct kset *queues_kset; 1980 #endif 1981 unsigned int promiscuity; 1982 unsigned int allmulti; 1983 1984 1985 /* Protocol-specific pointers */ 1986 1987 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1988 struct vlan_info __rcu *vlan_info; 1989 #endif 1990 #if IS_ENABLED(CONFIG_NET_DSA) 1991 struct dsa_port *dsa_ptr; 1992 #endif 1993 #if IS_ENABLED(CONFIG_TIPC) 1994 struct tipc_bearer __rcu *tipc_ptr; 1995 #endif 1996 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1997 void *atalk_ptr; 1998 #endif 1999 struct in_device __rcu *ip_ptr; 2000 #if IS_ENABLED(CONFIG_DECNET) 2001 struct dn_dev __rcu *dn_ptr; 2002 #endif 2003 struct inet6_dev __rcu *ip6_ptr; 2004 #if IS_ENABLED(CONFIG_AX25) 2005 void *ax25_ptr; 2006 #endif 2007 struct wireless_dev *ieee80211_ptr; 2008 struct wpan_dev *ieee802154_ptr; 2009 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2010 struct mpls_dev __rcu *mpls_ptr; 2011 #endif 2012 2013 /* 2014 * Cache lines mostly used on receive path (including eth_type_trans()) 2015 */ 2016 /* Interface address info used in eth_type_trans() */ 2017 unsigned char *dev_addr; 2018 2019 struct netdev_rx_queue *_rx; 2020 unsigned int num_rx_queues; 2021 unsigned int real_num_rx_queues; 2022 2023 struct bpf_prog __rcu *xdp_prog; 2024 unsigned long gro_flush_timeout; 2025 int napi_defer_hard_irqs; 2026 rx_handler_func_t __rcu *rx_handler; 2027 void __rcu *rx_handler_data; 2028 2029 #ifdef CONFIG_NET_CLS_ACT 2030 struct mini_Qdisc __rcu *miniq_ingress; 2031 #endif 2032 struct netdev_queue __rcu *ingress_queue; 2033 #ifdef CONFIG_NETFILTER_INGRESS 2034 struct nf_hook_entries __rcu *nf_hooks_ingress; 2035 #endif 2036 2037 unsigned char broadcast[MAX_ADDR_LEN]; 2038 #ifdef CONFIG_RFS_ACCEL 2039 struct cpu_rmap *rx_cpu_rmap; 2040 #endif 2041 struct hlist_node index_hlist; 2042 2043 /* 2044 * Cache lines mostly used on transmit path 2045 */ 2046 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2047 unsigned int num_tx_queues; 2048 unsigned int real_num_tx_queues; 2049 struct Qdisc *qdisc; 2050 unsigned int tx_queue_len; 2051 spinlock_t tx_global_lock; 2052 2053 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2054 2055 #ifdef CONFIG_XPS 2056 struct xps_dev_maps __rcu *xps_cpus_map; 2057 struct xps_dev_maps __rcu *xps_rxqs_map; 2058 #endif 2059 #ifdef CONFIG_NET_CLS_ACT 2060 struct mini_Qdisc __rcu *miniq_egress; 2061 #endif 2062 2063 #ifdef CONFIG_NET_SCHED 2064 DECLARE_HASHTABLE (qdisc_hash, 4); 2065 #endif 2066 /* These may be needed for future network-power-down code. */ 2067 struct timer_list watchdog_timer; 2068 int watchdog_timeo; 2069 2070 u32 proto_down_reason; 2071 2072 struct list_head todo_list; 2073 int __percpu *pcpu_refcnt; 2074 2075 struct list_head link_watch_list; 2076 2077 enum { NETREG_UNINITIALIZED=0, 2078 NETREG_REGISTERED, /* completed register_netdevice */ 2079 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2080 NETREG_UNREGISTERED, /* completed unregister todo */ 2081 NETREG_RELEASED, /* called free_netdev */ 2082 NETREG_DUMMY, /* dummy device for NAPI poll */ 2083 } reg_state:8; 2084 2085 bool dismantle; 2086 2087 enum { 2088 RTNL_LINK_INITIALIZED, 2089 RTNL_LINK_INITIALIZING, 2090 } rtnl_link_state:16; 2091 2092 bool needs_free_netdev; 2093 void (*priv_destructor)(struct net_device *dev); 2094 2095 #ifdef CONFIG_NETPOLL 2096 struct netpoll_info __rcu *npinfo; 2097 #endif 2098 2099 possible_net_t nd_net; 2100 2101 /* mid-layer private */ 2102 union { 2103 void *ml_priv; 2104 struct pcpu_lstats __percpu *lstats; 2105 struct pcpu_sw_netstats __percpu *tstats; 2106 struct pcpu_dstats __percpu *dstats; 2107 }; 2108 2109 #if IS_ENABLED(CONFIG_GARP) 2110 struct garp_port __rcu *garp_port; 2111 #endif 2112 #if IS_ENABLED(CONFIG_MRP) 2113 struct mrp_port __rcu *mrp_port; 2114 #endif 2115 2116 struct device dev; 2117 const struct attribute_group *sysfs_groups[4]; 2118 const struct attribute_group *sysfs_rx_queue_group; 2119 2120 const struct rtnl_link_ops *rtnl_link_ops; 2121 2122 /* for setting kernel sock attribute on TCP connection setup */ 2123 #define GSO_MAX_SIZE 65536 2124 unsigned int gso_max_size; 2125 #define GSO_MAX_SEGS 65535 2126 u16 gso_max_segs; 2127 2128 #ifdef CONFIG_DCB 2129 const struct dcbnl_rtnl_ops *dcbnl_ops; 2130 #endif 2131 s16 num_tc; 2132 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2133 u8 prio_tc_map[TC_BITMASK + 1]; 2134 2135 #if IS_ENABLED(CONFIG_FCOE) 2136 unsigned int fcoe_ddp_xid; 2137 #endif 2138 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2139 struct netprio_map __rcu *priomap; 2140 #endif 2141 struct phy_device *phydev; 2142 struct sfp_bus *sfp_bus; 2143 struct lock_class_key *qdisc_tx_busylock; 2144 struct lock_class_key *qdisc_running_key; 2145 bool proto_down; 2146 unsigned wol_enabled:1; 2147 2148 struct list_head net_notifier_list; 2149 2150 #if IS_ENABLED(CONFIG_MACSEC) 2151 /* MACsec management functions */ 2152 const struct macsec_ops *macsec_ops; 2153 #endif 2154 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2155 struct udp_tunnel_nic *udp_tunnel_nic; 2156 2157 /* protected by rtnl_lock */ 2158 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2159 }; 2160 #define to_net_dev(d) container_of(d, struct net_device, dev) 2161 2162 static inline bool netif_elide_gro(const struct net_device *dev) 2163 { 2164 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2165 return true; 2166 return false; 2167 } 2168 2169 #define NETDEV_ALIGN 32 2170 2171 static inline 2172 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2173 { 2174 return dev->prio_tc_map[prio & TC_BITMASK]; 2175 } 2176 2177 static inline 2178 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2179 { 2180 if (tc >= dev->num_tc) 2181 return -EINVAL; 2182 2183 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2184 return 0; 2185 } 2186 2187 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2188 void netdev_reset_tc(struct net_device *dev); 2189 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2190 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2191 2192 static inline 2193 int netdev_get_num_tc(struct net_device *dev) 2194 { 2195 return dev->num_tc; 2196 } 2197 2198 void netdev_unbind_sb_channel(struct net_device *dev, 2199 struct net_device *sb_dev); 2200 int netdev_bind_sb_channel_queue(struct net_device *dev, 2201 struct net_device *sb_dev, 2202 u8 tc, u16 count, u16 offset); 2203 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2204 static inline int netdev_get_sb_channel(struct net_device *dev) 2205 { 2206 return max_t(int, -dev->num_tc, 0); 2207 } 2208 2209 static inline 2210 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2211 unsigned int index) 2212 { 2213 return &dev->_tx[index]; 2214 } 2215 2216 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2217 const struct sk_buff *skb) 2218 { 2219 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2220 } 2221 2222 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2223 void (*f)(struct net_device *, 2224 struct netdev_queue *, 2225 void *), 2226 void *arg) 2227 { 2228 unsigned int i; 2229 2230 for (i = 0; i < dev->num_tx_queues; i++) 2231 f(dev, &dev->_tx[i], arg); 2232 } 2233 2234 #define netdev_lockdep_set_classes(dev) \ 2235 { \ 2236 static struct lock_class_key qdisc_tx_busylock_key; \ 2237 static struct lock_class_key qdisc_running_key; \ 2238 static struct lock_class_key qdisc_xmit_lock_key; \ 2239 static struct lock_class_key dev_addr_list_lock_key; \ 2240 unsigned int i; \ 2241 \ 2242 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2243 (dev)->qdisc_running_key = &qdisc_running_key; \ 2244 lockdep_set_class(&(dev)->addr_list_lock, \ 2245 &dev_addr_list_lock_key); \ 2246 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2247 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2248 &qdisc_xmit_lock_key); \ 2249 } 2250 2251 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2252 struct net_device *sb_dev); 2253 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2254 struct sk_buff *skb, 2255 struct net_device *sb_dev); 2256 2257 /* returns the headroom that the master device needs to take in account 2258 * when forwarding to this dev 2259 */ 2260 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2261 { 2262 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2263 } 2264 2265 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2266 { 2267 if (dev->netdev_ops->ndo_set_rx_headroom) 2268 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2269 } 2270 2271 /* set the device rx headroom to the dev's default */ 2272 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2273 { 2274 netdev_set_rx_headroom(dev, -1); 2275 } 2276 2277 /* 2278 * Net namespace inlines 2279 */ 2280 static inline 2281 struct net *dev_net(const struct net_device *dev) 2282 { 2283 return read_pnet(&dev->nd_net); 2284 } 2285 2286 static inline 2287 void dev_net_set(struct net_device *dev, struct net *net) 2288 { 2289 write_pnet(&dev->nd_net, net); 2290 } 2291 2292 /** 2293 * netdev_priv - access network device private data 2294 * @dev: network device 2295 * 2296 * Get network device private data 2297 */ 2298 static inline void *netdev_priv(const struct net_device *dev) 2299 { 2300 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2301 } 2302 2303 /* Set the sysfs physical device reference for the network logical device 2304 * if set prior to registration will cause a symlink during initialization. 2305 */ 2306 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2307 2308 /* Set the sysfs device type for the network logical device to allow 2309 * fine-grained identification of different network device types. For 2310 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2311 */ 2312 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2313 2314 /* Default NAPI poll() weight 2315 * Device drivers are strongly advised to not use bigger value 2316 */ 2317 #define NAPI_POLL_WEIGHT 64 2318 2319 /** 2320 * netif_napi_add - initialize a NAPI context 2321 * @dev: network device 2322 * @napi: NAPI context 2323 * @poll: polling function 2324 * @weight: default weight 2325 * 2326 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2327 * *any* of the other NAPI-related functions. 2328 */ 2329 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2330 int (*poll)(struct napi_struct *, int), int weight); 2331 2332 /** 2333 * netif_tx_napi_add - initialize a NAPI context 2334 * @dev: network device 2335 * @napi: NAPI context 2336 * @poll: polling function 2337 * @weight: default weight 2338 * 2339 * This variant of netif_napi_add() should be used from drivers using NAPI 2340 * to exclusively poll a TX queue. 2341 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2342 */ 2343 static inline void netif_tx_napi_add(struct net_device *dev, 2344 struct napi_struct *napi, 2345 int (*poll)(struct napi_struct *, int), 2346 int weight) 2347 { 2348 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2349 netif_napi_add(dev, napi, poll, weight); 2350 } 2351 2352 /** 2353 * netif_napi_del - remove a NAPI context 2354 * @napi: NAPI context 2355 * 2356 * netif_napi_del() removes a NAPI context from the network device NAPI list 2357 */ 2358 void netif_napi_del(struct napi_struct *napi); 2359 2360 struct napi_gro_cb { 2361 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2362 void *frag0; 2363 2364 /* Length of frag0. */ 2365 unsigned int frag0_len; 2366 2367 /* This indicates where we are processing relative to skb->data. */ 2368 int data_offset; 2369 2370 /* This is non-zero if the packet cannot be merged with the new skb. */ 2371 u16 flush; 2372 2373 /* Save the IP ID here and check when we get to the transport layer */ 2374 u16 flush_id; 2375 2376 /* Number of segments aggregated. */ 2377 u16 count; 2378 2379 /* Start offset for remote checksum offload */ 2380 u16 gro_remcsum_start; 2381 2382 /* jiffies when first packet was created/queued */ 2383 unsigned long age; 2384 2385 /* Used in ipv6_gro_receive() and foo-over-udp */ 2386 u16 proto; 2387 2388 /* This is non-zero if the packet may be of the same flow. */ 2389 u8 same_flow:1; 2390 2391 /* Used in tunnel GRO receive */ 2392 u8 encap_mark:1; 2393 2394 /* GRO checksum is valid */ 2395 u8 csum_valid:1; 2396 2397 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2398 u8 csum_cnt:3; 2399 2400 /* Free the skb? */ 2401 u8 free:2; 2402 #define NAPI_GRO_FREE 1 2403 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2404 2405 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2406 u8 is_ipv6:1; 2407 2408 /* Used in GRE, set in fou/gue_gro_receive */ 2409 u8 is_fou:1; 2410 2411 /* Used to determine if flush_id can be ignored */ 2412 u8 is_atomic:1; 2413 2414 /* Number of gro_receive callbacks this packet already went through */ 2415 u8 recursion_counter:4; 2416 2417 /* GRO is done by frag_list pointer chaining. */ 2418 u8 is_flist:1; 2419 2420 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2421 __wsum csum; 2422 2423 /* used in skb_gro_receive() slow path */ 2424 struct sk_buff *last; 2425 }; 2426 2427 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2428 2429 #define GRO_RECURSION_LIMIT 15 2430 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2431 { 2432 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2433 } 2434 2435 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2436 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2437 struct list_head *head, 2438 struct sk_buff *skb) 2439 { 2440 if (unlikely(gro_recursion_inc_test(skb))) { 2441 NAPI_GRO_CB(skb)->flush |= 1; 2442 return NULL; 2443 } 2444 2445 return cb(head, skb); 2446 } 2447 2448 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2449 struct sk_buff *); 2450 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2451 struct sock *sk, 2452 struct list_head *head, 2453 struct sk_buff *skb) 2454 { 2455 if (unlikely(gro_recursion_inc_test(skb))) { 2456 NAPI_GRO_CB(skb)->flush |= 1; 2457 return NULL; 2458 } 2459 2460 return cb(sk, head, skb); 2461 } 2462 2463 struct packet_type { 2464 __be16 type; /* This is really htons(ether_type). */ 2465 bool ignore_outgoing; 2466 struct net_device *dev; /* NULL is wildcarded here */ 2467 int (*func) (struct sk_buff *, 2468 struct net_device *, 2469 struct packet_type *, 2470 struct net_device *); 2471 void (*list_func) (struct list_head *, 2472 struct packet_type *, 2473 struct net_device *); 2474 bool (*id_match)(struct packet_type *ptype, 2475 struct sock *sk); 2476 void *af_packet_priv; 2477 struct list_head list; 2478 }; 2479 2480 struct offload_callbacks { 2481 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2482 netdev_features_t features); 2483 struct sk_buff *(*gro_receive)(struct list_head *head, 2484 struct sk_buff *skb); 2485 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2486 }; 2487 2488 struct packet_offload { 2489 __be16 type; /* This is really htons(ether_type). */ 2490 u16 priority; 2491 struct offload_callbacks callbacks; 2492 struct list_head list; 2493 }; 2494 2495 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2496 struct pcpu_sw_netstats { 2497 u64 rx_packets; 2498 u64 rx_bytes; 2499 u64 tx_packets; 2500 u64 tx_bytes; 2501 struct u64_stats_sync syncp; 2502 } __aligned(4 * sizeof(u64)); 2503 2504 struct pcpu_lstats { 2505 u64_stats_t packets; 2506 u64_stats_t bytes; 2507 struct u64_stats_sync syncp; 2508 } __aligned(2 * sizeof(u64)); 2509 2510 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2511 2512 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2513 { 2514 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2515 2516 u64_stats_update_begin(&lstats->syncp); 2517 u64_stats_add(&lstats->bytes, len); 2518 u64_stats_inc(&lstats->packets); 2519 u64_stats_update_end(&lstats->syncp); 2520 } 2521 2522 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2523 ({ \ 2524 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2525 if (pcpu_stats) { \ 2526 int __cpu; \ 2527 for_each_possible_cpu(__cpu) { \ 2528 typeof(type) *stat; \ 2529 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2530 u64_stats_init(&stat->syncp); \ 2531 } \ 2532 } \ 2533 pcpu_stats; \ 2534 }) 2535 2536 #define netdev_alloc_pcpu_stats(type) \ 2537 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2538 2539 enum netdev_lag_tx_type { 2540 NETDEV_LAG_TX_TYPE_UNKNOWN, 2541 NETDEV_LAG_TX_TYPE_RANDOM, 2542 NETDEV_LAG_TX_TYPE_BROADCAST, 2543 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2544 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2545 NETDEV_LAG_TX_TYPE_HASH, 2546 }; 2547 2548 enum netdev_lag_hash { 2549 NETDEV_LAG_HASH_NONE, 2550 NETDEV_LAG_HASH_L2, 2551 NETDEV_LAG_HASH_L34, 2552 NETDEV_LAG_HASH_L23, 2553 NETDEV_LAG_HASH_E23, 2554 NETDEV_LAG_HASH_E34, 2555 NETDEV_LAG_HASH_UNKNOWN, 2556 }; 2557 2558 struct netdev_lag_upper_info { 2559 enum netdev_lag_tx_type tx_type; 2560 enum netdev_lag_hash hash_type; 2561 }; 2562 2563 struct netdev_lag_lower_state_info { 2564 u8 link_up : 1, 2565 tx_enabled : 1; 2566 }; 2567 2568 #include <linux/notifier.h> 2569 2570 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2571 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2572 * adding new types. 2573 */ 2574 enum netdev_cmd { 2575 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2576 NETDEV_DOWN, 2577 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2578 detected a hardware crash and restarted 2579 - we can use this eg to kick tcp sessions 2580 once done */ 2581 NETDEV_CHANGE, /* Notify device state change */ 2582 NETDEV_REGISTER, 2583 NETDEV_UNREGISTER, 2584 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2585 NETDEV_CHANGEADDR, /* notify after the address change */ 2586 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2587 NETDEV_GOING_DOWN, 2588 NETDEV_CHANGENAME, 2589 NETDEV_FEAT_CHANGE, 2590 NETDEV_BONDING_FAILOVER, 2591 NETDEV_PRE_UP, 2592 NETDEV_PRE_TYPE_CHANGE, 2593 NETDEV_POST_TYPE_CHANGE, 2594 NETDEV_POST_INIT, 2595 NETDEV_RELEASE, 2596 NETDEV_NOTIFY_PEERS, 2597 NETDEV_JOIN, 2598 NETDEV_CHANGEUPPER, 2599 NETDEV_RESEND_IGMP, 2600 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2601 NETDEV_CHANGEINFODATA, 2602 NETDEV_BONDING_INFO, 2603 NETDEV_PRECHANGEUPPER, 2604 NETDEV_CHANGELOWERSTATE, 2605 NETDEV_UDP_TUNNEL_PUSH_INFO, 2606 NETDEV_UDP_TUNNEL_DROP_INFO, 2607 NETDEV_CHANGE_TX_QUEUE_LEN, 2608 NETDEV_CVLAN_FILTER_PUSH_INFO, 2609 NETDEV_CVLAN_FILTER_DROP_INFO, 2610 NETDEV_SVLAN_FILTER_PUSH_INFO, 2611 NETDEV_SVLAN_FILTER_DROP_INFO, 2612 }; 2613 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2614 2615 int register_netdevice_notifier(struct notifier_block *nb); 2616 int unregister_netdevice_notifier(struct notifier_block *nb); 2617 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2618 int unregister_netdevice_notifier_net(struct net *net, 2619 struct notifier_block *nb); 2620 int register_netdevice_notifier_dev_net(struct net_device *dev, 2621 struct notifier_block *nb, 2622 struct netdev_net_notifier *nn); 2623 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2624 struct notifier_block *nb, 2625 struct netdev_net_notifier *nn); 2626 2627 struct netdev_notifier_info { 2628 struct net_device *dev; 2629 struct netlink_ext_ack *extack; 2630 }; 2631 2632 struct netdev_notifier_info_ext { 2633 struct netdev_notifier_info info; /* must be first */ 2634 union { 2635 u32 mtu; 2636 } ext; 2637 }; 2638 2639 struct netdev_notifier_change_info { 2640 struct netdev_notifier_info info; /* must be first */ 2641 unsigned int flags_changed; 2642 }; 2643 2644 struct netdev_notifier_changeupper_info { 2645 struct netdev_notifier_info info; /* must be first */ 2646 struct net_device *upper_dev; /* new upper dev */ 2647 bool master; /* is upper dev master */ 2648 bool linking; /* is the notification for link or unlink */ 2649 void *upper_info; /* upper dev info */ 2650 }; 2651 2652 struct netdev_notifier_changelowerstate_info { 2653 struct netdev_notifier_info info; /* must be first */ 2654 void *lower_state_info; /* is lower dev state */ 2655 }; 2656 2657 struct netdev_notifier_pre_changeaddr_info { 2658 struct netdev_notifier_info info; /* must be first */ 2659 const unsigned char *dev_addr; 2660 }; 2661 2662 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2663 struct net_device *dev) 2664 { 2665 info->dev = dev; 2666 info->extack = NULL; 2667 } 2668 2669 static inline struct net_device * 2670 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2671 { 2672 return info->dev; 2673 } 2674 2675 static inline struct netlink_ext_ack * 2676 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2677 { 2678 return info->extack; 2679 } 2680 2681 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2682 2683 2684 extern rwlock_t dev_base_lock; /* Device list lock */ 2685 2686 #define for_each_netdev(net, d) \ 2687 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2688 #define for_each_netdev_reverse(net, d) \ 2689 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2690 #define for_each_netdev_rcu(net, d) \ 2691 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2692 #define for_each_netdev_safe(net, d, n) \ 2693 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2694 #define for_each_netdev_continue(net, d) \ 2695 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2696 #define for_each_netdev_continue_reverse(net, d) \ 2697 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2698 dev_list) 2699 #define for_each_netdev_continue_rcu(net, d) \ 2700 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2701 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2702 for_each_netdev_rcu(&init_net, slave) \ 2703 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2704 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2705 2706 static inline struct net_device *next_net_device(struct net_device *dev) 2707 { 2708 struct list_head *lh; 2709 struct net *net; 2710 2711 net = dev_net(dev); 2712 lh = dev->dev_list.next; 2713 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2714 } 2715 2716 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2717 { 2718 struct list_head *lh; 2719 struct net *net; 2720 2721 net = dev_net(dev); 2722 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2723 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2724 } 2725 2726 static inline struct net_device *first_net_device(struct net *net) 2727 { 2728 return list_empty(&net->dev_base_head) ? NULL : 2729 net_device_entry(net->dev_base_head.next); 2730 } 2731 2732 static inline struct net_device *first_net_device_rcu(struct net *net) 2733 { 2734 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2735 2736 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2737 } 2738 2739 int netdev_boot_setup_check(struct net_device *dev); 2740 unsigned long netdev_boot_base(const char *prefix, int unit); 2741 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2742 const char *hwaddr); 2743 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2744 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2745 void dev_add_pack(struct packet_type *pt); 2746 void dev_remove_pack(struct packet_type *pt); 2747 void __dev_remove_pack(struct packet_type *pt); 2748 void dev_add_offload(struct packet_offload *po); 2749 void dev_remove_offload(struct packet_offload *po); 2750 2751 int dev_get_iflink(const struct net_device *dev); 2752 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2753 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2754 unsigned short mask); 2755 struct net_device *dev_get_by_name(struct net *net, const char *name); 2756 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2757 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2758 int dev_alloc_name(struct net_device *dev, const char *name); 2759 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2760 void dev_close(struct net_device *dev); 2761 void dev_close_many(struct list_head *head, bool unlink); 2762 void dev_disable_lro(struct net_device *dev); 2763 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2764 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2765 struct net_device *sb_dev); 2766 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2767 struct net_device *sb_dev); 2768 int dev_queue_xmit(struct sk_buff *skb); 2769 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2770 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2771 int register_netdevice(struct net_device *dev); 2772 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2773 void unregister_netdevice_many(struct list_head *head); 2774 static inline void unregister_netdevice(struct net_device *dev) 2775 { 2776 unregister_netdevice_queue(dev, NULL); 2777 } 2778 2779 int netdev_refcnt_read(const struct net_device *dev); 2780 void free_netdev(struct net_device *dev); 2781 void netdev_freemem(struct net_device *dev); 2782 void synchronize_net(void); 2783 int init_dummy_netdev(struct net_device *dev); 2784 2785 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2786 struct sk_buff *skb, 2787 bool all_slaves); 2788 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2789 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2790 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2791 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2792 int netdev_get_name(struct net *net, char *name, int ifindex); 2793 int dev_restart(struct net_device *dev); 2794 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2795 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2796 2797 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2798 { 2799 return NAPI_GRO_CB(skb)->data_offset; 2800 } 2801 2802 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2803 { 2804 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2805 } 2806 2807 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2808 { 2809 NAPI_GRO_CB(skb)->data_offset += len; 2810 } 2811 2812 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2813 unsigned int offset) 2814 { 2815 return NAPI_GRO_CB(skb)->frag0 + offset; 2816 } 2817 2818 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2819 { 2820 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2821 } 2822 2823 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2824 { 2825 NAPI_GRO_CB(skb)->frag0 = NULL; 2826 NAPI_GRO_CB(skb)->frag0_len = 0; 2827 } 2828 2829 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2830 unsigned int offset) 2831 { 2832 if (!pskb_may_pull(skb, hlen)) 2833 return NULL; 2834 2835 skb_gro_frag0_invalidate(skb); 2836 return skb->data + offset; 2837 } 2838 2839 static inline void *skb_gro_network_header(struct sk_buff *skb) 2840 { 2841 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2842 skb_network_offset(skb); 2843 } 2844 2845 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2846 const void *start, unsigned int len) 2847 { 2848 if (NAPI_GRO_CB(skb)->csum_valid) 2849 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2850 csum_partial(start, len, 0)); 2851 } 2852 2853 /* GRO checksum functions. These are logical equivalents of the normal 2854 * checksum functions (in skbuff.h) except that they operate on the GRO 2855 * offsets and fields in sk_buff. 2856 */ 2857 2858 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2859 2860 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2861 { 2862 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2863 } 2864 2865 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2866 bool zero_okay, 2867 __sum16 check) 2868 { 2869 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2870 skb_checksum_start_offset(skb) < 2871 skb_gro_offset(skb)) && 2872 !skb_at_gro_remcsum_start(skb) && 2873 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2874 (!zero_okay || check)); 2875 } 2876 2877 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2878 __wsum psum) 2879 { 2880 if (NAPI_GRO_CB(skb)->csum_valid && 2881 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2882 return 0; 2883 2884 NAPI_GRO_CB(skb)->csum = psum; 2885 2886 return __skb_gro_checksum_complete(skb); 2887 } 2888 2889 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2890 { 2891 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2892 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2893 NAPI_GRO_CB(skb)->csum_cnt--; 2894 } else { 2895 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2896 * verified a new top level checksum or an encapsulated one 2897 * during GRO. This saves work if we fallback to normal path. 2898 */ 2899 __skb_incr_checksum_unnecessary(skb); 2900 } 2901 } 2902 2903 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2904 compute_pseudo) \ 2905 ({ \ 2906 __sum16 __ret = 0; \ 2907 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2908 __ret = __skb_gro_checksum_validate_complete(skb, \ 2909 compute_pseudo(skb, proto)); \ 2910 if (!__ret) \ 2911 skb_gro_incr_csum_unnecessary(skb); \ 2912 __ret; \ 2913 }) 2914 2915 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2916 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2917 2918 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2919 compute_pseudo) \ 2920 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2921 2922 #define skb_gro_checksum_simple_validate(skb) \ 2923 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2924 2925 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2926 { 2927 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2928 !NAPI_GRO_CB(skb)->csum_valid); 2929 } 2930 2931 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2932 __wsum pseudo) 2933 { 2934 NAPI_GRO_CB(skb)->csum = ~pseudo; 2935 NAPI_GRO_CB(skb)->csum_valid = 1; 2936 } 2937 2938 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2939 do { \ 2940 if (__skb_gro_checksum_convert_check(skb)) \ 2941 __skb_gro_checksum_convert(skb, \ 2942 compute_pseudo(skb, proto)); \ 2943 } while (0) 2944 2945 struct gro_remcsum { 2946 int offset; 2947 __wsum delta; 2948 }; 2949 2950 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2951 { 2952 grc->offset = 0; 2953 grc->delta = 0; 2954 } 2955 2956 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2957 unsigned int off, size_t hdrlen, 2958 int start, int offset, 2959 struct gro_remcsum *grc, 2960 bool nopartial) 2961 { 2962 __wsum delta; 2963 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2964 2965 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2966 2967 if (!nopartial) { 2968 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2969 return ptr; 2970 } 2971 2972 ptr = skb_gro_header_fast(skb, off); 2973 if (skb_gro_header_hard(skb, off + plen)) { 2974 ptr = skb_gro_header_slow(skb, off + plen, off); 2975 if (!ptr) 2976 return NULL; 2977 } 2978 2979 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2980 start, offset); 2981 2982 /* Adjust skb->csum since we changed the packet */ 2983 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2984 2985 grc->offset = off + hdrlen + offset; 2986 grc->delta = delta; 2987 2988 return ptr; 2989 } 2990 2991 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2992 struct gro_remcsum *grc) 2993 { 2994 void *ptr; 2995 size_t plen = grc->offset + sizeof(u16); 2996 2997 if (!grc->delta) 2998 return; 2999 3000 ptr = skb_gro_header_fast(skb, grc->offset); 3001 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3002 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3003 if (!ptr) 3004 return; 3005 } 3006 3007 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3008 } 3009 3010 #ifdef CONFIG_XFRM_OFFLOAD 3011 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3012 { 3013 if (PTR_ERR(pp) != -EINPROGRESS) 3014 NAPI_GRO_CB(skb)->flush |= flush; 3015 } 3016 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3017 struct sk_buff *pp, 3018 int flush, 3019 struct gro_remcsum *grc) 3020 { 3021 if (PTR_ERR(pp) != -EINPROGRESS) { 3022 NAPI_GRO_CB(skb)->flush |= flush; 3023 skb_gro_remcsum_cleanup(skb, grc); 3024 skb->remcsum_offload = 0; 3025 } 3026 } 3027 #else 3028 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3029 { 3030 NAPI_GRO_CB(skb)->flush |= flush; 3031 } 3032 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3033 struct sk_buff *pp, 3034 int flush, 3035 struct gro_remcsum *grc) 3036 { 3037 NAPI_GRO_CB(skb)->flush |= flush; 3038 skb_gro_remcsum_cleanup(skb, grc); 3039 skb->remcsum_offload = 0; 3040 } 3041 #endif 3042 3043 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3044 unsigned short type, 3045 const void *daddr, const void *saddr, 3046 unsigned int len) 3047 { 3048 if (!dev->header_ops || !dev->header_ops->create) 3049 return 0; 3050 3051 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3052 } 3053 3054 static inline int dev_parse_header(const struct sk_buff *skb, 3055 unsigned char *haddr) 3056 { 3057 const struct net_device *dev = skb->dev; 3058 3059 if (!dev->header_ops || !dev->header_ops->parse) 3060 return 0; 3061 return dev->header_ops->parse(skb, haddr); 3062 } 3063 3064 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3065 { 3066 const struct net_device *dev = skb->dev; 3067 3068 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3069 return 0; 3070 return dev->header_ops->parse_protocol(skb); 3071 } 3072 3073 /* ll_header must have at least hard_header_len allocated */ 3074 static inline bool dev_validate_header(const struct net_device *dev, 3075 char *ll_header, int len) 3076 { 3077 if (likely(len >= dev->hard_header_len)) 3078 return true; 3079 if (len < dev->min_header_len) 3080 return false; 3081 3082 if (capable(CAP_SYS_RAWIO)) { 3083 memset(ll_header + len, 0, dev->hard_header_len - len); 3084 return true; 3085 } 3086 3087 if (dev->header_ops && dev->header_ops->validate) 3088 return dev->header_ops->validate(ll_header, len); 3089 3090 return false; 3091 } 3092 3093 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3094 int len, int size); 3095 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3096 static inline int unregister_gifconf(unsigned int family) 3097 { 3098 return register_gifconf(family, NULL); 3099 } 3100 3101 #ifdef CONFIG_NET_FLOW_LIMIT 3102 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3103 struct sd_flow_limit { 3104 u64 count; 3105 unsigned int num_buckets; 3106 unsigned int history_head; 3107 u16 history[FLOW_LIMIT_HISTORY]; 3108 u8 buckets[]; 3109 }; 3110 3111 extern int netdev_flow_limit_table_len; 3112 #endif /* CONFIG_NET_FLOW_LIMIT */ 3113 3114 /* 3115 * Incoming packets are placed on per-CPU queues 3116 */ 3117 struct softnet_data { 3118 struct list_head poll_list; 3119 struct sk_buff_head process_queue; 3120 3121 /* stats */ 3122 unsigned int processed; 3123 unsigned int time_squeeze; 3124 unsigned int received_rps; 3125 #ifdef CONFIG_RPS 3126 struct softnet_data *rps_ipi_list; 3127 #endif 3128 #ifdef CONFIG_NET_FLOW_LIMIT 3129 struct sd_flow_limit __rcu *flow_limit; 3130 #endif 3131 struct Qdisc *output_queue; 3132 struct Qdisc **output_queue_tailp; 3133 struct sk_buff *completion_queue; 3134 #ifdef CONFIG_XFRM_OFFLOAD 3135 struct sk_buff_head xfrm_backlog; 3136 #endif 3137 /* written and read only by owning cpu: */ 3138 struct { 3139 u16 recursion; 3140 u8 more; 3141 } xmit; 3142 #ifdef CONFIG_RPS 3143 /* input_queue_head should be written by cpu owning this struct, 3144 * and only read by other cpus. Worth using a cache line. 3145 */ 3146 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3147 3148 /* Elements below can be accessed between CPUs for RPS/RFS */ 3149 call_single_data_t csd ____cacheline_aligned_in_smp; 3150 struct softnet_data *rps_ipi_next; 3151 unsigned int cpu; 3152 unsigned int input_queue_tail; 3153 #endif 3154 unsigned int dropped; 3155 struct sk_buff_head input_pkt_queue; 3156 struct napi_struct backlog; 3157 3158 }; 3159 3160 static inline void input_queue_head_incr(struct softnet_data *sd) 3161 { 3162 #ifdef CONFIG_RPS 3163 sd->input_queue_head++; 3164 #endif 3165 } 3166 3167 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3168 unsigned int *qtail) 3169 { 3170 #ifdef CONFIG_RPS 3171 *qtail = ++sd->input_queue_tail; 3172 #endif 3173 } 3174 3175 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3176 3177 static inline int dev_recursion_level(void) 3178 { 3179 return this_cpu_read(softnet_data.xmit.recursion); 3180 } 3181 3182 #define XMIT_RECURSION_LIMIT 8 3183 static inline bool dev_xmit_recursion(void) 3184 { 3185 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3186 XMIT_RECURSION_LIMIT); 3187 } 3188 3189 static inline void dev_xmit_recursion_inc(void) 3190 { 3191 __this_cpu_inc(softnet_data.xmit.recursion); 3192 } 3193 3194 static inline void dev_xmit_recursion_dec(void) 3195 { 3196 __this_cpu_dec(softnet_data.xmit.recursion); 3197 } 3198 3199 void __netif_schedule(struct Qdisc *q); 3200 void netif_schedule_queue(struct netdev_queue *txq); 3201 3202 static inline void netif_tx_schedule_all(struct net_device *dev) 3203 { 3204 unsigned int i; 3205 3206 for (i = 0; i < dev->num_tx_queues; i++) 3207 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3208 } 3209 3210 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3211 { 3212 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3213 } 3214 3215 /** 3216 * netif_start_queue - allow transmit 3217 * @dev: network device 3218 * 3219 * Allow upper layers to call the device hard_start_xmit routine. 3220 */ 3221 static inline void netif_start_queue(struct net_device *dev) 3222 { 3223 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3224 } 3225 3226 static inline void netif_tx_start_all_queues(struct net_device *dev) 3227 { 3228 unsigned int i; 3229 3230 for (i = 0; i < dev->num_tx_queues; i++) { 3231 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3232 netif_tx_start_queue(txq); 3233 } 3234 } 3235 3236 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3237 3238 /** 3239 * netif_wake_queue - restart transmit 3240 * @dev: network device 3241 * 3242 * Allow upper layers to call the device hard_start_xmit routine. 3243 * Used for flow control when transmit resources are available. 3244 */ 3245 static inline void netif_wake_queue(struct net_device *dev) 3246 { 3247 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3248 } 3249 3250 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3251 { 3252 unsigned int i; 3253 3254 for (i = 0; i < dev->num_tx_queues; i++) { 3255 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3256 netif_tx_wake_queue(txq); 3257 } 3258 } 3259 3260 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3261 { 3262 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3263 } 3264 3265 /** 3266 * netif_stop_queue - stop transmitted packets 3267 * @dev: network device 3268 * 3269 * Stop upper layers calling the device hard_start_xmit routine. 3270 * Used for flow control when transmit resources are unavailable. 3271 */ 3272 static inline void netif_stop_queue(struct net_device *dev) 3273 { 3274 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3275 } 3276 3277 void netif_tx_stop_all_queues(struct net_device *dev); 3278 3279 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3280 { 3281 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3282 } 3283 3284 /** 3285 * netif_queue_stopped - test if transmit queue is flowblocked 3286 * @dev: network device 3287 * 3288 * Test if transmit queue on device is currently unable to send. 3289 */ 3290 static inline bool netif_queue_stopped(const struct net_device *dev) 3291 { 3292 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3293 } 3294 3295 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3296 { 3297 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3298 } 3299 3300 static inline bool 3301 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3302 { 3303 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3304 } 3305 3306 static inline bool 3307 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3308 { 3309 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3310 } 3311 3312 /** 3313 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3314 * @dev_queue: pointer to transmit queue 3315 * 3316 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3317 * to give appropriate hint to the CPU. 3318 */ 3319 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3320 { 3321 #ifdef CONFIG_BQL 3322 prefetchw(&dev_queue->dql.num_queued); 3323 #endif 3324 } 3325 3326 /** 3327 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3328 * @dev_queue: pointer to transmit queue 3329 * 3330 * BQL enabled drivers might use this helper in their TX completion path, 3331 * to give appropriate hint to the CPU. 3332 */ 3333 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3334 { 3335 #ifdef CONFIG_BQL 3336 prefetchw(&dev_queue->dql.limit); 3337 #endif 3338 } 3339 3340 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3341 unsigned int bytes) 3342 { 3343 #ifdef CONFIG_BQL 3344 dql_queued(&dev_queue->dql, bytes); 3345 3346 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3347 return; 3348 3349 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3350 3351 /* 3352 * The XOFF flag must be set before checking the dql_avail below, 3353 * because in netdev_tx_completed_queue we update the dql_completed 3354 * before checking the XOFF flag. 3355 */ 3356 smp_mb(); 3357 3358 /* check again in case another CPU has just made room avail */ 3359 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3360 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3361 #endif 3362 } 3363 3364 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3365 * that they should not test BQL status themselves. 3366 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3367 * skb of a batch. 3368 * Returns true if the doorbell must be used to kick the NIC. 3369 */ 3370 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3371 unsigned int bytes, 3372 bool xmit_more) 3373 { 3374 if (xmit_more) { 3375 #ifdef CONFIG_BQL 3376 dql_queued(&dev_queue->dql, bytes); 3377 #endif 3378 return netif_tx_queue_stopped(dev_queue); 3379 } 3380 netdev_tx_sent_queue(dev_queue, bytes); 3381 return true; 3382 } 3383 3384 /** 3385 * netdev_sent_queue - report the number of bytes queued to hardware 3386 * @dev: network device 3387 * @bytes: number of bytes queued to the hardware device queue 3388 * 3389 * Report the number of bytes queued for sending/completion to the network 3390 * device hardware queue. @bytes should be a good approximation and should 3391 * exactly match netdev_completed_queue() @bytes 3392 */ 3393 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3394 { 3395 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3396 } 3397 3398 static inline bool __netdev_sent_queue(struct net_device *dev, 3399 unsigned int bytes, 3400 bool xmit_more) 3401 { 3402 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3403 xmit_more); 3404 } 3405 3406 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3407 unsigned int pkts, unsigned int bytes) 3408 { 3409 #ifdef CONFIG_BQL 3410 if (unlikely(!bytes)) 3411 return; 3412 3413 dql_completed(&dev_queue->dql, bytes); 3414 3415 /* 3416 * Without the memory barrier there is a small possiblity that 3417 * netdev_tx_sent_queue will miss the update and cause the queue to 3418 * be stopped forever 3419 */ 3420 smp_mb(); 3421 3422 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3423 return; 3424 3425 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3426 netif_schedule_queue(dev_queue); 3427 #endif 3428 } 3429 3430 /** 3431 * netdev_completed_queue - report bytes and packets completed by device 3432 * @dev: network device 3433 * @pkts: actual number of packets sent over the medium 3434 * @bytes: actual number of bytes sent over the medium 3435 * 3436 * Report the number of bytes and packets transmitted by the network device 3437 * hardware queue over the physical medium, @bytes must exactly match the 3438 * @bytes amount passed to netdev_sent_queue() 3439 */ 3440 static inline void netdev_completed_queue(struct net_device *dev, 3441 unsigned int pkts, unsigned int bytes) 3442 { 3443 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3444 } 3445 3446 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3447 { 3448 #ifdef CONFIG_BQL 3449 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3450 dql_reset(&q->dql); 3451 #endif 3452 } 3453 3454 /** 3455 * netdev_reset_queue - reset the packets and bytes count of a network device 3456 * @dev_queue: network device 3457 * 3458 * Reset the bytes and packet count of a network device and clear the 3459 * software flow control OFF bit for this network device 3460 */ 3461 static inline void netdev_reset_queue(struct net_device *dev_queue) 3462 { 3463 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3464 } 3465 3466 /** 3467 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3468 * @dev: network device 3469 * @queue_index: given tx queue index 3470 * 3471 * Returns 0 if given tx queue index >= number of device tx queues, 3472 * otherwise returns the originally passed tx queue index. 3473 */ 3474 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3475 { 3476 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3477 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3478 dev->name, queue_index, 3479 dev->real_num_tx_queues); 3480 return 0; 3481 } 3482 3483 return queue_index; 3484 } 3485 3486 /** 3487 * netif_running - test if up 3488 * @dev: network device 3489 * 3490 * Test if the device has been brought up. 3491 */ 3492 static inline bool netif_running(const struct net_device *dev) 3493 { 3494 return test_bit(__LINK_STATE_START, &dev->state); 3495 } 3496 3497 /* 3498 * Routines to manage the subqueues on a device. We only need start, 3499 * stop, and a check if it's stopped. All other device management is 3500 * done at the overall netdevice level. 3501 * Also test the device if we're multiqueue. 3502 */ 3503 3504 /** 3505 * netif_start_subqueue - allow sending packets on subqueue 3506 * @dev: network device 3507 * @queue_index: sub queue index 3508 * 3509 * Start individual transmit queue of a device with multiple transmit queues. 3510 */ 3511 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3512 { 3513 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3514 3515 netif_tx_start_queue(txq); 3516 } 3517 3518 /** 3519 * netif_stop_subqueue - stop sending packets on subqueue 3520 * @dev: network device 3521 * @queue_index: sub queue index 3522 * 3523 * Stop individual transmit queue of a device with multiple transmit queues. 3524 */ 3525 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3526 { 3527 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3528 netif_tx_stop_queue(txq); 3529 } 3530 3531 /** 3532 * netif_subqueue_stopped - test status of subqueue 3533 * @dev: network device 3534 * @queue_index: sub queue index 3535 * 3536 * Check individual transmit queue of a device with multiple transmit queues. 3537 */ 3538 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3539 u16 queue_index) 3540 { 3541 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3542 3543 return netif_tx_queue_stopped(txq); 3544 } 3545 3546 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3547 struct sk_buff *skb) 3548 { 3549 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3550 } 3551 3552 /** 3553 * netif_wake_subqueue - allow sending packets on subqueue 3554 * @dev: network device 3555 * @queue_index: sub queue index 3556 * 3557 * Resume individual transmit queue of a device with multiple transmit queues. 3558 */ 3559 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3560 { 3561 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3562 3563 netif_tx_wake_queue(txq); 3564 } 3565 3566 #ifdef CONFIG_XPS 3567 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3568 u16 index); 3569 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3570 u16 index, bool is_rxqs_map); 3571 3572 /** 3573 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3574 * @j: CPU/Rx queue index 3575 * @mask: bitmask of all cpus/rx queues 3576 * @nr_bits: number of bits in the bitmask 3577 * 3578 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3579 */ 3580 static inline bool netif_attr_test_mask(unsigned long j, 3581 const unsigned long *mask, 3582 unsigned int nr_bits) 3583 { 3584 cpu_max_bits_warn(j, nr_bits); 3585 return test_bit(j, mask); 3586 } 3587 3588 /** 3589 * netif_attr_test_online - Test for online CPU/Rx queue 3590 * @j: CPU/Rx queue index 3591 * @online_mask: bitmask for CPUs/Rx queues that are online 3592 * @nr_bits: number of bits in the bitmask 3593 * 3594 * Returns true if a CPU/Rx queue is online. 3595 */ 3596 static inline bool netif_attr_test_online(unsigned long j, 3597 const unsigned long *online_mask, 3598 unsigned int nr_bits) 3599 { 3600 cpu_max_bits_warn(j, nr_bits); 3601 3602 if (online_mask) 3603 return test_bit(j, online_mask); 3604 3605 return (j < nr_bits); 3606 } 3607 3608 /** 3609 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3610 * @n: CPU/Rx queue index 3611 * @srcp: the cpumask/Rx queue mask pointer 3612 * @nr_bits: number of bits in the bitmask 3613 * 3614 * Returns >= nr_bits if no further CPUs/Rx queues set. 3615 */ 3616 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3617 unsigned int nr_bits) 3618 { 3619 /* -1 is a legal arg here. */ 3620 if (n != -1) 3621 cpu_max_bits_warn(n, nr_bits); 3622 3623 if (srcp) 3624 return find_next_bit(srcp, nr_bits, n + 1); 3625 3626 return n + 1; 3627 } 3628 3629 /** 3630 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3631 * @n: CPU/Rx queue index 3632 * @src1p: the first CPUs/Rx queues mask pointer 3633 * @src2p: the second CPUs/Rx queues mask pointer 3634 * @nr_bits: number of bits in the bitmask 3635 * 3636 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3637 */ 3638 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3639 const unsigned long *src2p, 3640 unsigned int nr_bits) 3641 { 3642 /* -1 is a legal arg here. */ 3643 if (n != -1) 3644 cpu_max_bits_warn(n, nr_bits); 3645 3646 if (src1p && src2p) 3647 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3648 else if (src1p) 3649 return find_next_bit(src1p, nr_bits, n + 1); 3650 else if (src2p) 3651 return find_next_bit(src2p, nr_bits, n + 1); 3652 3653 return n + 1; 3654 } 3655 #else 3656 static inline int netif_set_xps_queue(struct net_device *dev, 3657 const struct cpumask *mask, 3658 u16 index) 3659 { 3660 return 0; 3661 } 3662 3663 static inline int __netif_set_xps_queue(struct net_device *dev, 3664 const unsigned long *mask, 3665 u16 index, bool is_rxqs_map) 3666 { 3667 return 0; 3668 } 3669 #endif 3670 3671 /** 3672 * netif_is_multiqueue - test if device has multiple transmit queues 3673 * @dev: network device 3674 * 3675 * Check if device has multiple transmit queues 3676 */ 3677 static inline bool netif_is_multiqueue(const struct net_device *dev) 3678 { 3679 return dev->num_tx_queues > 1; 3680 } 3681 3682 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3683 3684 #ifdef CONFIG_SYSFS 3685 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3686 #else 3687 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3688 unsigned int rxqs) 3689 { 3690 dev->real_num_rx_queues = rxqs; 3691 return 0; 3692 } 3693 #endif 3694 3695 static inline struct netdev_rx_queue * 3696 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3697 { 3698 return dev->_rx + rxq; 3699 } 3700 3701 #ifdef CONFIG_SYSFS 3702 static inline unsigned int get_netdev_rx_queue_index( 3703 struct netdev_rx_queue *queue) 3704 { 3705 struct net_device *dev = queue->dev; 3706 int index = queue - dev->_rx; 3707 3708 BUG_ON(index >= dev->num_rx_queues); 3709 return index; 3710 } 3711 #endif 3712 3713 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3714 int netif_get_num_default_rss_queues(void); 3715 3716 enum skb_free_reason { 3717 SKB_REASON_CONSUMED, 3718 SKB_REASON_DROPPED, 3719 }; 3720 3721 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3722 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3723 3724 /* 3725 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3726 * interrupt context or with hardware interrupts being disabled. 3727 * (in_irq() || irqs_disabled()) 3728 * 3729 * We provide four helpers that can be used in following contexts : 3730 * 3731 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3732 * replacing kfree_skb(skb) 3733 * 3734 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3735 * Typically used in place of consume_skb(skb) in TX completion path 3736 * 3737 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3738 * replacing kfree_skb(skb) 3739 * 3740 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3741 * and consumed a packet. Used in place of consume_skb(skb) 3742 */ 3743 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3744 { 3745 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3746 } 3747 3748 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3749 { 3750 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3751 } 3752 3753 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3754 { 3755 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3756 } 3757 3758 static inline void dev_consume_skb_any(struct sk_buff *skb) 3759 { 3760 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3761 } 3762 3763 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3764 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3765 int netif_rx(struct sk_buff *skb); 3766 int netif_rx_ni(struct sk_buff *skb); 3767 int netif_receive_skb(struct sk_buff *skb); 3768 int netif_receive_skb_core(struct sk_buff *skb); 3769 void netif_receive_skb_list(struct list_head *head); 3770 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3771 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3772 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3773 gro_result_t napi_gro_frags(struct napi_struct *napi); 3774 struct packet_offload *gro_find_receive_by_type(__be16 type); 3775 struct packet_offload *gro_find_complete_by_type(__be16 type); 3776 3777 static inline void napi_free_frags(struct napi_struct *napi) 3778 { 3779 kfree_skb(napi->skb); 3780 napi->skb = NULL; 3781 } 3782 3783 bool netdev_is_rx_handler_busy(struct net_device *dev); 3784 int netdev_rx_handler_register(struct net_device *dev, 3785 rx_handler_func_t *rx_handler, 3786 void *rx_handler_data); 3787 void netdev_rx_handler_unregister(struct net_device *dev); 3788 3789 bool dev_valid_name(const char *name); 3790 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3791 bool *need_copyout); 3792 int dev_ifconf(struct net *net, struct ifconf *, int); 3793 int dev_ethtool(struct net *net, struct ifreq *); 3794 unsigned int dev_get_flags(const struct net_device *); 3795 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3796 struct netlink_ext_ack *extack); 3797 int dev_change_flags(struct net_device *dev, unsigned int flags, 3798 struct netlink_ext_ack *extack); 3799 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3800 unsigned int gchanges); 3801 int dev_change_name(struct net_device *, const char *); 3802 int dev_set_alias(struct net_device *, const char *, size_t); 3803 int dev_get_alias(const struct net_device *, char *, size_t); 3804 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3805 int __dev_set_mtu(struct net_device *, int); 3806 int dev_validate_mtu(struct net_device *dev, int mtu, 3807 struct netlink_ext_ack *extack); 3808 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3809 struct netlink_ext_ack *extack); 3810 int dev_set_mtu(struct net_device *, int); 3811 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3812 void dev_set_group(struct net_device *, int); 3813 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3814 struct netlink_ext_ack *extack); 3815 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3816 struct netlink_ext_ack *extack); 3817 int dev_change_carrier(struct net_device *, bool new_carrier); 3818 int dev_get_phys_port_id(struct net_device *dev, 3819 struct netdev_phys_item_id *ppid); 3820 int dev_get_phys_port_name(struct net_device *dev, 3821 char *name, size_t len); 3822 int dev_get_port_parent_id(struct net_device *dev, 3823 struct netdev_phys_item_id *ppid, bool recurse); 3824 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3825 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3826 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3827 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3828 u32 value); 3829 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3830 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3831 struct netdev_queue *txq, int *ret); 3832 3833 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3834 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3835 int fd, int expected_fd, u32 flags); 3836 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3837 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3838 3839 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3840 3841 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3842 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3843 bool is_skb_forwardable(const struct net_device *dev, 3844 const struct sk_buff *skb); 3845 3846 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3847 struct sk_buff *skb) 3848 { 3849 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3850 unlikely(!is_skb_forwardable(dev, skb))) { 3851 atomic_long_inc(&dev->rx_dropped); 3852 kfree_skb(skb); 3853 return NET_RX_DROP; 3854 } 3855 3856 skb_scrub_packet(skb, true); 3857 skb->priority = 0; 3858 return 0; 3859 } 3860 3861 bool dev_nit_active(struct net_device *dev); 3862 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3863 3864 extern int netdev_budget; 3865 extern unsigned int netdev_budget_usecs; 3866 3867 /* Called by rtnetlink.c:rtnl_unlock() */ 3868 void netdev_run_todo(void); 3869 3870 /** 3871 * dev_put - release reference to device 3872 * @dev: network device 3873 * 3874 * Release reference to device to allow it to be freed. 3875 */ 3876 static inline void dev_put(struct net_device *dev) 3877 { 3878 this_cpu_dec(*dev->pcpu_refcnt); 3879 } 3880 3881 /** 3882 * dev_hold - get reference to device 3883 * @dev: network device 3884 * 3885 * Hold reference to device to keep it from being freed. 3886 */ 3887 static inline void dev_hold(struct net_device *dev) 3888 { 3889 this_cpu_inc(*dev->pcpu_refcnt); 3890 } 3891 3892 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3893 * and _off may be called from IRQ context, but it is caller 3894 * who is responsible for serialization of these calls. 3895 * 3896 * The name carrier is inappropriate, these functions should really be 3897 * called netif_lowerlayer_*() because they represent the state of any 3898 * kind of lower layer not just hardware media. 3899 */ 3900 3901 void linkwatch_init_dev(struct net_device *dev); 3902 void linkwatch_fire_event(struct net_device *dev); 3903 void linkwatch_forget_dev(struct net_device *dev); 3904 3905 /** 3906 * netif_carrier_ok - test if carrier present 3907 * @dev: network device 3908 * 3909 * Check if carrier is present on device 3910 */ 3911 static inline bool netif_carrier_ok(const struct net_device *dev) 3912 { 3913 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3914 } 3915 3916 unsigned long dev_trans_start(struct net_device *dev); 3917 3918 void __netdev_watchdog_up(struct net_device *dev); 3919 3920 void netif_carrier_on(struct net_device *dev); 3921 3922 void netif_carrier_off(struct net_device *dev); 3923 3924 /** 3925 * netif_dormant_on - mark device as dormant. 3926 * @dev: network device 3927 * 3928 * Mark device as dormant (as per RFC2863). 3929 * 3930 * The dormant state indicates that the relevant interface is not 3931 * actually in a condition to pass packets (i.e., it is not 'up') but is 3932 * in a "pending" state, waiting for some external event. For "on- 3933 * demand" interfaces, this new state identifies the situation where the 3934 * interface is waiting for events to place it in the up state. 3935 */ 3936 static inline void netif_dormant_on(struct net_device *dev) 3937 { 3938 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3939 linkwatch_fire_event(dev); 3940 } 3941 3942 /** 3943 * netif_dormant_off - set device as not dormant. 3944 * @dev: network device 3945 * 3946 * Device is not in dormant state. 3947 */ 3948 static inline void netif_dormant_off(struct net_device *dev) 3949 { 3950 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3951 linkwatch_fire_event(dev); 3952 } 3953 3954 /** 3955 * netif_dormant - test if device is dormant 3956 * @dev: network device 3957 * 3958 * Check if device is dormant. 3959 */ 3960 static inline bool netif_dormant(const struct net_device *dev) 3961 { 3962 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3963 } 3964 3965 3966 /** 3967 * netif_testing_on - mark device as under test. 3968 * @dev: network device 3969 * 3970 * Mark device as under test (as per RFC2863). 3971 * 3972 * The testing state indicates that some test(s) must be performed on 3973 * the interface. After completion, of the test, the interface state 3974 * will change to up, dormant, or down, as appropriate. 3975 */ 3976 static inline void netif_testing_on(struct net_device *dev) 3977 { 3978 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 3979 linkwatch_fire_event(dev); 3980 } 3981 3982 /** 3983 * netif_testing_off - set device as not under test. 3984 * @dev: network device 3985 * 3986 * Device is not in testing state. 3987 */ 3988 static inline void netif_testing_off(struct net_device *dev) 3989 { 3990 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 3991 linkwatch_fire_event(dev); 3992 } 3993 3994 /** 3995 * netif_testing - test if device is under test 3996 * @dev: network device 3997 * 3998 * Check if device is under test 3999 */ 4000 static inline bool netif_testing(const struct net_device *dev) 4001 { 4002 return test_bit(__LINK_STATE_TESTING, &dev->state); 4003 } 4004 4005 4006 /** 4007 * netif_oper_up - test if device is operational 4008 * @dev: network device 4009 * 4010 * Check if carrier is operational 4011 */ 4012 static inline bool netif_oper_up(const struct net_device *dev) 4013 { 4014 return (dev->operstate == IF_OPER_UP || 4015 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4016 } 4017 4018 /** 4019 * netif_device_present - is device available or removed 4020 * @dev: network device 4021 * 4022 * Check if device has not been removed from system. 4023 */ 4024 static inline bool netif_device_present(struct net_device *dev) 4025 { 4026 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4027 } 4028 4029 void netif_device_detach(struct net_device *dev); 4030 4031 void netif_device_attach(struct net_device *dev); 4032 4033 /* 4034 * Network interface message level settings 4035 */ 4036 4037 enum { 4038 NETIF_MSG_DRV_BIT, 4039 NETIF_MSG_PROBE_BIT, 4040 NETIF_MSG_LINK_BIT, 4041 NETIF_MSG_TIMER_BIT, 4042 NETIF_MSG_IFDOWN_BIT, 4043 NETIF_MSG_IFUP_BIT, 4044 NETIF_MSG_RX_ERR_BIT, 4045 NETIF_MSG_TX_ERR_BIT, 4046 NETIF_MSG_TX_QUEUED_BIT, 4047 NETIF_MSG_INTR_BIT, 4048 NETIF_MSG_TX_DONE_BIT, 4049 NETIF_MSG_RX_STATUS_BIT, 4050 NETIF_MSG_PKTDATA_BIT, 4051 NETIF_MSG_HW_BIT, 4052 NETIF_MSG_WOL_BIT, 4053 4054 /* When you add a new bit above, update netif_msg_class_names array 4055 * in net/ethtool/common.c 4056 */ 4057 NETIF_MSG_CLASS_COUNT, 4058 }; 4059 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4060 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4061 4062 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4063 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4064 4065 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4066 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4067 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4068 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4069 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4070 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4071 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4072 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4073 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4074 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4075 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4076 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4077 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4078 #define NETIF_MSG_HW __NETIF_MSG(HW) 4079 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4080 4081 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4082 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4083 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4084 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4085 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4086 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4087 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4088 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4089 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4090 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4091 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4092 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4093 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4094 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4095 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4096 4097 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4098 { 4099 /* use default */ 4100 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4101 return default_msg_enable_bits; 4102 if (debug_value == 0) /* no output */ 4103 return 0; 4104 /* set low N bits */ 4105 return (1U << debug_value) - 1; 4106 } 4107 4108 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4109 { 4110 spin_lock(&txq->_xmit_lock); 4111 txq->xmit_lock_owner = cpu; 4112 } 4113 4114 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4115 { 4116 __acquire(&txq->_xmit_lock); 4117 return true; 4118 } 4119 4120 static inline void __netif_tx_release(struct netdev_queue *txq) 4121 { 4122 __release(&txq->_xmit_lock); 4123 } 4124 4125 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4126 { 4127 spin_lock_bh(&txq->_xmit_lock); 4128 txq->xmit_lock_owner = smp_processor_id(); 4129 } 4130 4131 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4132 { 4133 bool ok = spin_trylock(&txq->_xmit_lock); 4134 if (likely(ok)) 4135 txq->xmit_lock_owner = smp_processor_id(); 4136 return ok; 4137 } 4138 4139 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4140 { 4141 txq->xmit_lock_owner = -1; 4142 spin_unlock(&txq->_xmit_lock); 4143 } 4144 4145 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4146 { 4147 txq->xmit_lock_owner = -1; 4148 spin_unlock_bh(&txq->_xmit_lock); 4149 } 4150 4151 static inline void txq_trans_update(struct netdev_queue *txq) 4152 { 4153 if (txq->xmit_lock_owner != -1) 4154 txq->trans_start = jiffies; 4155 } 4156 4157 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4158 static inline void netif_trans_update(struct net_device *dev) 4159 { 4160 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4161 4162 if (txq->trans_start != jiffies) 4163 txq->trans_start = jiffies; 4164 } 4165 4166 /** 4167 * netif_tx_lock - grab network device transmit lock 4168 * @dev: network device 4169 * 4170 * Get network device transmit lock 4171 */ 4172 static inline void netif_tx_lock(struct net_device *dev) 4173 { 4174 unsigned int i; 4175 int cpu; 4176 4177 spin_lock(&dev->tx_global_lock); 4178 cpu = smp_processor_id(); 4179 for (i = 0; i < dev->num_tx_queues; i++) { 4180 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4181 4182 /* We are the only thread of execution doing a 4183 * freeze, but we have to grab the _xmit_lock in 4184 * order to synchronize with threads which are in 4185 * the ->hard_start_xmit() handler and already 4186 * checked the frozen bit. 4187 */ 4188 __netif_tx_lock(txq, cpu); 4189 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4190 __netif_tx_unlock(txq); 4191 } 4192 } 4193 4194 static inline void netif_tx_lock_bh(struct net_device *dev) 4195 { 4196 local_bh_disable(); 4197 netif_tx_lock(dev); 4198 } 4199 4200 static inline void netif_tx_unlock(struct net_device *dev) 4201 { 4202 unsigned int i; 4203 4204 for (i = 0; i < dev->num_tx_queues; i++) { 4205 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4206 4207 /* No need to grab the _xmit_lock here. If the 4208 * queue is not stopped for another reason, we 4209 * force a schedule. 4210 */ 4211 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4212 netif_schedule_queue(txq); 4213 } 4214 spin_unlock(&dev->tx_global_lock); 4215 } 4216 4217 static inline void netif_tx_unlock_bh(struct net_device *dev) 4218 { 4219 netif_tx_unlock(dev); 4220 local_bh_enable(); 4221 } 4222 4223 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4224 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4225 __netif_tx_lock(txq, cpu); \ 4226 } else { \ 4227 __netif_tx_acquire(txq); \ 4228 } \ 4229 } 4230 4231 #define HARD_TX_TRYLOCK(dev, txq) \ 4232 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4233 __netif_tx_trylock(txq) : \ 4234 __netif_tx_acquire(txq)) 4235 4236 #define HARD_TX_UNLOCK(dev, txq) { \ 4237 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4238 __netif_tx_unlock(txq); \ 4239 } else { \ 4240 __netif_tx_release(txq); \ 4241 } \ 4242 } 4243 4244 static inline void netif_tx_disable(struct net_device *dev) 4245 { 4246 unsigned int i; 4247 int cpu; 4248 4249 local_bh_disable(); 4250 cpu = smp_processor_id(); 4251 for (i = 0; i < dev->num_tx_queues; i++) { 4252 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4253 4254 __netif_tx_lock(txq, cpu); 4255 netif_tx_stop_queue(txq); 4256 __netif_tx_unlock(txq); 4257 } 4258 local_bh_enable(); 4259 } 4260 4261 static inline void netif_addr_lock(struct net_device *dev) 4262 { 4263 spin_lock(&dev->addr_list_lock); 4264 } 4265 4266 static inline void netif_addr_lock_nested(struct net_device *dev) 4267 { 4268 spin_lock_nested(&dev->addr_list_lock, dev->lower_level); 4269 } 4270 4271 static inline void netif_addr_lock_bh(struct net_device *dev) 4272 { 4273 spin_lock_bh(&dev->addr_list_lock); 4274 } 4275 4276 static inline void netif_addr_unlock(struct net_device *dev) 4277 { 4278 spin_unlock(&dev->addr_list_lock); 4279 } 4280 4281 static inline void netif_addr_unlock_bh(struct net_device *dev) 4282 { 4283 spin_unlock_bh(&dev->addr_list_lock); 4284 } 4285 4286 /* 4287 * dev_addrs walker. Should be used only for read access. Call with 4288 * rcu_read_lock held. 4289 */ 4290 #define for_each_dev_addr(dev, ha) \ 4291 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4292 4293 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4294 4295 void ether_setup(struct net_device *dev); 4296 4297 /* Support for loadable net-drivers */ 4298 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4299 unsigned char name_assign_type, 4300 void (*setup)(struct net_device *), 4301 unsigned int txqs, unsigned int rxqs); 4302 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4303 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4304 4305 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4306 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4307 count) 4308 4309 int register_netdev(struct net_device *dev); 4310 void unregister_netdev(struct net_device *dev); 4311 4312 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4313 4314 /* General hardware address lists handling functions */ 4315 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4316 struct netdev_hw_addr_list *from_list, int addr_len); 4317 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4318 struct netdev_hw_addr_list *from_list, int addr_len); 4319 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4320 struct net_device *dev, 4321 int (*sync)(struct net_device *, const unsigned char *), 4322 int (*unsync)(struct net_device *, 4323 const unsigned char *)); 4324 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4325 struct net_device *dev, 4326 int (*sync)(struct net_device *, 4327 const unsigned char *, int), 4328 int (*unsync)(struct net_device *, 4329 const unsigned char *, int)); 4330 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4331 struct net_device *dev, 4332 int (*unsync)(struct net_device *, 4333 const unsigned char *, int)); 4334 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4335 struct net_device *dev, 4336 int (*unsync)(struct net_device *, 4337 const unsigned char *)); 4338 void __hw_addr_init(struct netdev_hw_addr_list *list); 4339 4340 /* Functions used for device addresses handling */ 4341 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4342 unsigned char addr_type); 4343 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4344 unsigned char addr_type); 4345 void dev_addr_flush(struct net_device *dev); 4346 int dev_addr_init(struct net_device *dev); 4347 4348 /* Functions used for unicast addresses handling */ 4349 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4350 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4351 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4352 int dev_uc_sync(struct net_device *to, struct net_device *from); 4353 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4354 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4355 void dev_uc_flush(struct net_device *dev); 4356 void dev_uc_init(struct net_device *dev); 4357 4358 /** 4359 * __dev_uc_sync - Synchonize device's unicast list 4360 * @dev: device to sync 4361 * @sync: function to call if address should be added 4362 * @unsync: function to call if address should be removed 4363 * 4364 * Add newly added addresses to the interface, and release 4365 * addresses that have been deleted. 4366 */ 4367 static inline int __dev_uc_sync(struct net_device *dev, 4368 int (*sync)(struct net_device *, 4369 const unsigned char *), 4370 int (*unsync)(struct net_device *, 4371 const unsigned char *)) 4372 { 4373 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4374 } 4375 4376 /** 4377 * __dev_uc_unsync - Remove synchronized addresses from device 4378 * @dev: device to sync 4379 * @unsync: function to call if address should be removed 4380 * 4381 * Remove all addresses that were added to the device by dev_uc_sync(). 4382 */ 4383 static inline void __dev_uc_unsync(struct net_device *dev, 4384 int (*unsync)(struct net_device *, 4385 const unsigned char *)) 4386 { 4387 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4388 } 4389 4390 /* Functions used for multicast addresses handling */ 4391 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4392 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4393 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4394 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4395 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4396 int dev_mc_sync(struct net_device *to, struct net_device *from); 4397 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4398 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4399 void dev_mc_flush(struct net_device *dev); 4400 void dev_mc_init(struct net_device *dev); 4401 4402 /** 4403 * __dev_mc_sync - Synchonize device's multicast list 4404 * @dev: device to sync 4405 * @sync: function to call if address should be added 4406 * @unsync: function to call if address should be removed 4407 * 4408 * Add newly added addresses to the interface, and release 4409 * addresses that have been deleted. 4410 */ 4411 static inline int __dev_mc_sync(struct net_device *dev, 4412 int (*sync)(struct net_device *, 4413 const unsigned char *), 4414 int (*unsync)(struct net_device *, 4415 const unsigned char *)) 4416 { 4417 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4418 } 4419 4420 /** 4421 * __dev_mc_unsync - Remove synchronized addresses from device 4422 * @dev: device to sync 4423 * @unsync: function to call if address should be removed 4424 * 4425 * Remove all addresses that were added to the device by dev_mc_sync(). 4426 */ 4427 static inline void __dev_mc_unsync(struct net_device *dev, 4428 int (*unsync)(struct net_device *, 4429 const unsigned char *)) 4430 { 4431 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4432 } 4433 4434 /* Functions used for secondary unicast and multicast support */ 4435 void dev_set_rx_mode(struct net_device *dev); 4436 void __dev_set_rx_mode(struct net_device *dev); 4437 int dev_set_promiscuity(struct net_device *dev, int inc); 4438 int dev_set_allmulti(struct net_device *dev, int inc); 4439 void netdev_state_change(struct net_device *dev); 4440 void netdev_notify_peers(struct net_device *dev); 4441 void netdev_features_change(struct net_device *dev); 4442 /* Load a device via the kmod */ 4443 void dev_load(struct net *net, const char *name); 4444 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4445 struct rtnl_link_stats64 *storage); 4446 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4447 const struct net_device_stats *netdev_stats); 4448 4449 extern int netdev_max_backlog; 4450 extern int netdev_tstamp_prequeue; 4451 extern int weight_p; 4452 extern int dev_weight_rx_bias; 4453 extern int dev_weight_tx_bias; 4454 extern int dev_rx_weight; 4455 extern int dev_tx_weight; 4456 extern int gro_normal_batch; 4457 4458 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4459 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4460 struct list_head **iter); 4461 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4462 struct list_head **iter); 4463 4464 /* iterate through upper list, must be called under RCU read lock */ 4465 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4466 for (iter = &(dev)->adj_list.upper, \ 4467 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4468 updev; \ 4469 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4470 4471 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4472 int (*fn)(struct net_device *upper_dev, 4473 void *data), 4474 void *data); 4475 4476 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4477 struct net_device *upper_dev); 4478 4479 bool netdev_has_any_upper_dev(struct net_device *dev); 4480 4481 void *netdev_lower_get_next_private(struct net_device *dev, 4482 struct list_head **iter); 4483 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4484 struct list_head **iter); 4485 4486 #define netdev_for_each_lower_private(dev, priv, iter) \ 4487 for (iter = (dev)->adj_list.lower.next, \ 4488 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4489 priv; \ 4490 priv = netdev_lower_get_next_private(dev, &(iter))) 4491 4492 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4493 for (iter = &(dev)->adj_list.lower, \ 4494 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4495 priv; \ 4496 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4497 4498 void *netdev_lower_get_next(struct net_device *dev, 4499 struct list_head **iter); 4500 4501 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4502 for (iter = (dev)->adj_list.lower.next, \ 4503 ldev = netdev_lower_get_next(dev, &(iter)); \ 4504 ldev; \ 4505 ldev = netdev_lower_get_next(dev, &(iter))) 4506 4507 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4508 struct list_head **iter); 4509 int netdev_walk_all_lower_dev(struct net_device *dev, 4510 int (*fn)(struct net_device *lower_dev, 4511 void *data), 4512 void *data); 4513 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4514 int (*fn)(struct net_device *lower_dev, 4515 void *data), 4516 void *data); 4517 4518 void *netdev_adjacent_get_private(struct list_head *adj_list); 4519 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4520 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4521 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4522 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4523 struct netlink_ext_ack *extack); 4524 int netdev_master_upper_dev_link(struct net_device *dev, 4525 struct net_device *upper_dev, 4526 void *upper_priv, void *upper_info, 4527 struct netlink_ext_ack *extack); 4528 void netdev_upper_dev_unlink(struct net_device *dev, 4529 struct net_device *upper_dev); 4530 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4531 struct net_device *new_dev, 4532 struct net_device *dev, 4533 struct netlink_ext_ack *extack); 4534 void netdev_adjacent_change_commit(struct net_device *old_dev, 4535 struct net_device *new_dev, 4536 struct net_device *dev); 4537 void netdev_adjacent_change_abort(struct net_device *old_dev, 4538 struct net_device *new_dev, 4539 struct net_device *dev); 4540 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4541 void *netdev_lower_dev_get_private(struct net_device *dev, 4542 struct net_device *lower_dev); 4543 void netdev_lower_state_changed(struct net_device *lower_dev, 4544 void *lower_state_info); 4545 4546 /* RSS keys are 40 or 52 bytes long */ 4547 #define NETDEV_RSS_KEY_LEN 52 4548 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4549 void netdev_rss_key_fill(void *buffer, size_t len); 4550 4551 int skb_checksum_help(struct sk_buff *skb); 4552 int skb_crc32c_csum_help(struct sk_buff *skb); 4553 int skb_csum_hwoffload_help(struct sk_buff *skb, 4554 const netdev_features_t features); 4555 4556 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4557 netdev_features_t features, bool tx_path); 4558 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4559 netdev_features_t features); 4560 4561 struct netdev_bonding_info { 4562 ifslave slave; 4563 ifbond master; 4564 }; 4565 4566 struct netdev_notifier_bonding_info { 4567 struct netdev_notifier_info info; /* must be first */ 4568 struct netdev_bonding_info bonding_info; 4569 }; 4570 4571 void netdev_bonding_info_change(struct net_device *dev, 4572 struct netdev_bonding_info *bonding_info); 4573 4574 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4575 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4576 #else 4577 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4578 const void *data) 4579 { 4580 } 4581 #endif 4582 4583 static inline 4584 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4585 { 4586 return __skb_gso_segment(skb, features, true); 4587 } 4588 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4589 4590 static inline bool can_checksum_protocol(netdev_features_t features, 4591 __be16 protocol) 4592 { 4593 if (protocol == htons(ETH_P_FCOE)) 4594 return !!(features & NETIF_F_FCOE_CRC); 4595 4596 /* Assume this is an IP checksum (not SCTP CRC) */ 4597 4598 if (features & NETIF_F_HW_CSUM) { 4599 /* Can checksum everything */ 4600 return true; 4601 } 4602 4603 switch (protocol) { 4604 case htons(ETH_P_IP): 4605 return !!(features & NETIF_F_IP_CSUM); 4606 case htons(ETH_P_IPV6): 4607 return !!(features & NETIF_F_IPV6_CSUM); 4608 default: 4609 return false; 4610 } 4611 } 4612 4613 #ifdef CONFIG_BUG 4614 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4615 #else 4616 static inline void netdev_rx_csum_fault(struct net_device *dev, 4617 struct sk_buff *skb) 4618 { 4619 } 4620 #endif 4621 /* rx skb timestamps */ 4622 void net_enable_timestamp(void); 4623 void net_disable_timestamp(void); 4624 4625 #ifdef CONFIG_PROC_FS 4626 int __init dev_proc_init(void); 4627 #else 4628 #define dev_proc_init() 0 4629 #endif 4630 4631 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4632 struct sk_buff *skb, struct net_device *dev, 4633 bool more) 4634 { 4635 __this_cpu_write(softnet_data.xmit.more, more); 4636 return ops->ndo_start_xmit(skb, dev); 4637 } 4638 4639 static inline bool netdev_xmit_more(void) 4640 { 4641 return __this_cpu_read(softnet_data.xmit.more); 4642 } 4643 4644 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4645 struct netdev_queue *txq, bool more) 4646 { 4647 const struct net_device_ops *ops = dev->netdev_ops; 4648 netdev_tx_t rc; 4649 4650 rc = __netdev_start_xmit(ops, skb, dev, more); 4651 if (rc == NETDEV_TX_OK) 4652 txq_trans_update(txq); 4653 4654 return rc; 4655 } 4656 4657 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4658 const void *ns); 4659 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4660 const void *ns); 4661 4662 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4663 { 4664 return netdev_class_create_file_ns(class_attr, NULL); 4665 } 4666 4667 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4668 { 4669 netdev_class_remove_file_ns(class_attr, NULL); 4670 } 4671 4672 extern const struct kobj_ns_type_operations net_ns_type_operations; 4673 4674 const char *netdev_drivername(const struct net_device *dev); 4675 4676 void linkwatch_run_queue(void); 4677 4678 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4679 netdev_features_t f2) 4680 { 4681 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4682 if (f1 & NETIF_F_HW_CSUM) 4683 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4684 else 4685 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4686 } 4687 4688 return f1 & f2; 4689 } 4690 4691 static inline netdev_features_t netdev_get_wanted_features( 4692 struct net_device *dev) 4693 { 4694 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4695 } 4696 netdev_features_t netdev_increment_features(netdev_features_t all, 4697 netdev_features_t one, netdev_features_t mask); 4698 4699 /* Allow TSO being used on stacked device : 4700 * Performing the GSO segmentation before last device 4701 * is a performance improvement. 4702 */ 4703 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4704 netdev_features_t mask) 4705 { 4706 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4707 } 4708 4709 int __netdev_update_features(struct net_device *dev); 4710 void netdev_update_features(struct net_device *dev); 4711 void netdev_change_features(struct net_device *dev); 4712 4713 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4714 struct net_device *dev); 4715 4716 netdev_features_t passthru_features_check(struct sk_buff *skb, 4717 struct net_device *dev, 4718 netdev_features_t features); 4719 netdev_features_t netif_skb_features(struct sk_buff *skb); 4720 4721 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4722 { 4723 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4724 4725 /* check flags correspondence */ 4726 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4727 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4728 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4729 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4730 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4731 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4732 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4733 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4734 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4735 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4736 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4737 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4738 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4739 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4740 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4741 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4742 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4743 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4744 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4745 4746 return (features & feature) == feature; 4747 } 4748 4749 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4750 { 4751 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4752 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4753 } 4754 4755 static inline bool netif_needs_gso(struct sk_buff *skb, 4756 netdev_features_t features) 4757 { 4758 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4759 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4760 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4761 } 4762 4763 static inline void netif_set_gso_max_size(struct net_device *dev, 4764 unsigned int size) 4765 { 4766 dev->gso_max_size = size; 4767 } 4768 4769 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4770 int pulled_hlen, u16 mac_offset, 4771 int mac_len) 4772 { 4773 skb->protocol = protocol; 4774 skb->encapsulation = 1; 4775 skb_push(skb, pulled_hlen); 4776 skb_reset_transport_header(skb); 4777 skb->mac_header = mac_offset; 4778 skb->network_header = skb->mac_header + mac_len; 4779 skb->mac_len = mac_len; 4780 } 4781 4782 static inline bool netif_is_macsec(const struct net_device *dev) 4783 { 4784 return dev->priv_flags & IFF_MACSEC; 4785 } 4786 4787 static inline bool netif_is_macvlan(const struct net_device *dev) 4788 { 4789 return dev->priv_flags & IFF_MACVLAN; 4790 } 4791 4792 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4793 { 4794 return dev->priv_flags & IFF_MACVLAN_PORT; 4795 } 4796 4797 static inline bool netif_is_bond_master(const struct net_device *dev) 4798 { 4799 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4800 } 4801 4802 static inline bool netif_is_bond_slave(const struct net_device *dev) 4803 { 4804 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4805 } 4806 4807 static inline bool netif_supports_nofcs(struct net_device *dev) 4808 { 4809 return dev->priv_flags & IFF_SUPP_NOFCS; 4810 } 4811 4812 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4813 { 4814 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4815 } 4816 4817 static inline bool netif_is_l3_master(const struct net_device *dev) 4818 { 4819 return dev->priv_flags & IFF_L3MDEV_MASTER; 4820 } 4821 4822 static inline bool netif_is_l3_slave(const struct net_device *dev) 4823 { 4824 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4825 } 4826 4827 static inline bool netif_is_bridge_master(const struct net_device *dev) 4828 { 4829 return dev->priv_flags & IFF_EBRIDGE; 4830 } 4831 4832 static inline bool netif_is_bridge_port(const struct net_device *dev) 4833 { 4834 return dev->priv_flags & IFF_BRIDGE_PORT; 4835 } 4836 4837 static inline bool netif_is_ovs_master(const struct net_device *dev) 4838 { 4839 return dev->priv_flags & IFF_OPENVSWITCH; 4840 } 4841 4842 static inline bool netif_is_ovs_port(const struct net_device *dev) 4843 { 4844 return dev->priv_flags & IFF_OVS_DATAPATH; 4845 } 4846 4847 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4848 { 4849 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4850 } 4851 4852 static inline bool netif_is_team_master(const struct net_device *dev) 4853 { 4854 return dev->priv_flags & IFF_TEAM; 4855 } 4856 4857 static inline bool netif_is_team_port(const struct net_device *dev) 4858 { 4859 return dev->priv_flags & IFF_TEAM_PORT; 4860 } 4861 4862 static inline bool netif_is_lag_master(const struct net_device *dev) 4863 { 4864 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4865 } 4866 4867 static inline bool netif_is_lag_port(const struct net_device *dev) 4868 { 4869 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4870 } 4871 4872 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4873 { 4874 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4875 } 4876 4877 static inline bool netif_is_failover(const struct net_device *dev) 4878 { 4879 return dev->priv_flags & IFF_FAILOVER; 4880 } 4881 4882 static inline bool netif_is_failover_slave(const struct net_device *dev) 4883 { 4884 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4885 } 4886 4887 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4888 static inline void netif_keep_dst(struct net_device *dev) 4889 { 4890 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4891 } 4892 4893 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4894 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4895 { 4896 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4897 return dev->priv_flags & IFF_MACSEC; 4898 } 4899 4900 extern struct pernet_operations __net_initdata loopback_net_ops; 4901 4902 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4903 4904 /* netdev_printk helpers, similar to dev_printk */ 4905 4906 static inline const char *netdev_name(const struct net_device *dev) 4907 { 4908 if (!dev->name[0] || strchr(dev->name, '%')) 4909 return "(unnamed net_device)"; 4910 return dev->name; 4911 } 4912 4913 static inline bool netdev_unregistering(const struct net_device *dev) 4914 { 4915 return dev->reg_state == NETREG_UNREGISTERING; 4916 } 4917 4918 static inline const char *netdev_reg_state(const struct net_device *dev) 4919 { 4920 switch (dev->reg_state) { 4921 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4922 case NETREG_REGISTERED: return ""; 4923 case NETREG_UNREGISTERING: return " (unregistering)"; 4924 case NETREG_UNREGISTERED: return " (unregistered)"; 4925 case NETREG_RELEASED: return " (released)"; 4926 case NETREG_DUMMY: return " (dummy)"; 4927 } 4928 4929 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4930 return " (unknown)"; 4931 } 4932 4933 __printf(3, 4) __cold 4934 void netdev_printk(const char *level, const struct net_device *dev, 4935 const char *format, ...); 4936 __printf(2, 3) __cold 4937 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4938 __printf(2, 3) __cold 4939 void netdev_alert(const struct net_device *dev, const char *format, ...); 4940 __printf(2, 3) __cold 4941 void netdev_crit(const struct net_device *dev, const char *format, ...); 4942 __printf(2, 3) __cold 4943 void netdev_err(const struct net_device *dev, const char *format, ...); 4944 __printf(2, 3) __cold 4945 void netdev_warn(const struct net_device *dev, const char *format, ...); 4946 __printf(2, 3) __cold 4947 void netdev_notice(const struct net_device *dev, const char *format, ...); 4948 __printf(2, 3) __cold 4949 void netdev_info(const struct net_device *dev, const char *format, ...); 4950 4951 #define netdev_level_once(level, dev, fmt, ...) \ 4952 do { \ 4953 static bool __print_once __read_mostly; \ 4954 \ 4955 if (!__print_once) { \ 4956 __print_once = true; \ 4957 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4958 } \ 4959 } while (0) 4960 4961 #define netdev_emerg_once(dev, fmt, ...) \ 4962 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4963 #define netdev_alert_once(dev, fmt, ...) \ 4964 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4965 #define netdev_crit_once(dev, fmt, ...) \ 4966 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4967 #define netdev_err_once(dev, fmt, ...) \ 4968 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4969 #define netdev_warn_once(dev, fmt, ...) \ 4970 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4971 #define netdev_notice_once(dev, fmt, ...) \ 4972 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4973 #define netdev_info_once(dev, fmt, ...) \ 4974 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4975 4976 #define MODULE_ALIAS_NETDEV(device) \ 4977 MODULE_ALIAS("netdev-" device) 4978 4979 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 4980 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 4981 #define netdev_dbg(__dev, format, args...) \ 4982 do { \ 4983 dynamic_netdev_dbg(__dev, format, ##args); \ 4984 } while (0) 4985 #elif defined(DEBUG) 4986 #define netdev_dbg(__dev, format, args...) \ 4987 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4988 #else 4989 #define netdev_dbg(__dev, format, args...) \ 4990 ({ \ 4991 if (0) \ 4992 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4993 }) 4994 #endif 4995 4996 #if defined(VERBOSE_DEBUG) 4997 #define netdev_vdbg netdev_dbg 4998 #else 4999 5000 #define netdev_vdbg(dev, format, args...) \ 5001 ({ \ 5002 if (0) \ 5003 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5004 0; \ 5005 }) 5006 #endif 5007 5008 /* 5009 * netdev_WARN() acts like dev_printk(), but with the key difference 5010 * of using a WARN/WARN_ON to get the message out, including the 5011 * file/line information and a backtrace. 5012 */ 5013 #define netdev_WARN(dev, format, args...) \ 5014 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5015 netdev_reg_state(dev), ##args) 5016 5017 #define netdev_WARN_ONCE(dev, format, args...) \ 5018 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5019 netdev_reg_state(dev), ##args) 5020 5021 /* netif printk helpers, similar to netdev_printk */ 5022 5023 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5024 do { \ 5025 if (netif_msg_##type(priv)) \ 5026 netdev_printk(level, (dev), fmt, ##args); \ 5027 } while (0) 5028 5029 #define netif_level(level, priv, type, dev, fmt, args...) \ 5030 do { \ 5031 if (netif_msg_##type(priv)) \ 5032 netdev_##level(dev, fmt, ##args); \ 5033 } while (0) 5034 5035 #define netif_emerg(priv, type, dev, fmt, args...) \ 5036 netif_level(emerg, priv, type, dev, fmt, ##args) 5037 #define netif_alert(priv, type, dev, fmt, args...) \ 5038 netif_level(alert, priv, type, dev, fmt, ##args) 5039 #define netif_crit(priv, type, dev, fmt, args...) \ 5040 netif_level(crit, priv, type, dev, fmt, ##args) 5041 #define netif_err(priv, type, dev, fmt, args...) \ 5042 netif_level(err, priv, type, dev, fmt, ##args) 5043 #define netif_warn(priv, type, dev, fmt, args...) \ 5044 netif_level(warn, priv, type, dev, fmt, ##args) 5045 #define netif_notice(priv, type, dev, fmt, args...) \ 5046 netif_level(notice, priv, type, dev, fmt, ##args) 5047 #define netif_info(priv, type, dev, fmt, args...) \ 5048 netif_level(info, priv, type, dev, fmt, ##args) 5049 5050 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5051 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5052 #define netif_dbg(priv, type, netdev, format, args...) \ 5053 do { \ 5054 if (netif_msg_##type(priv)) \ 5055 dynamic_netdev_dbg(netdev, format, ##args); \ 5056 } while (0) 5057 #elif defined(DEBUG) 5058 #define netif_dbg(priv, type, dev, format, args...) \ 5059 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5060 #else 5061 #define netif_dbg(priv, type, dev, format, args...) \ 5062 ({ \ 5063 if (0) \ 5064 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5065 0; \ 5066 }) 5067 #endif 5068 5069 /* if @cond then downgrade to debug, else print at @level */ 5070 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5071 do { \ 5072 if (cond) \ 5073 netif_dbg(priv, type, netdev, fmt, ##args); \ 5074 else \ 5075 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5076 } while (0) 5077 5078 #if defined(VERBOSE_DEBUG) 5079 #define netif_vdbg netif_dbg 5080 #else 5081 #define netif_vdbg(priv, type, dev, format, args...) \ 5082 ({ \ 5083 if (0) \ 5084 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5085 0; \ 5086 }) 5087 #endif 5088 5089 /* 5090 * The list of packet types we will receive (as opposed to discard) 5091 * and the routines to invoke. 5092 * 5093 * Why 16. Because with 16 the only overlap we get on a hash of the 5094 * low nibble of the protocol value is RARP/SNAP/X.25. 5095 * 5096 * 0800 IP 5097 * 0001 802.3 5098 * 0002 AX.25 5099 * 0004 802.2 5100 * 8035 RARP 5101 * 0005 SNAP 5102 * 0805 X.25 5103 * 0806 ARP 5104 * 8137 IPX 5105 * 0009 Localtalk 5106 * 86DD IPv6 5107 */ 5108 #define PTYPE_HASH_SIZE (16) 5109 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5110 5111 extern struct net_device *blackhole_netdev; 5112 5113 #endif /* _LINUX_NETDEVICE_H */ 5114